Slot measuring and calculating method and device for resource management products and electronic equipment

Abstract

The invention provides a storage space measuring and calculating method and device for resource management products and electronic equipment. The method comprises the following steps: acquiring an industry space measurement and calculation request of a target resource management product, wherein the industry space measurement and calculation request comprises target industry information and measurement and calculation time information; obtaining first historical income statistical data of the target resource management product and obtaining second historical income statistical data corresponding to the target industry information; determining a plurality of measurement and calculation data acquisition time points based on the measurement and calculation time information, and determining a revenue relation data sequence corresponding to the measurement and calculation time information according to the plurality of measurement and calculation data acquisition time points, the first historical revenue statistical data and the second historical revenue statistical data; and calling a target neural network model to carry out bin position measurement and calculation on the income relation data sequence to obtain bin position information corresponding to the target industry in the target resource management product corresponding to the measurement and calculation time information. According to the method, the calculation accuracy of the industry shipping space information corresponding to the resource management product can be improved.

Description

Technical Field

[0001] The present disclosure relates to the field of data processing technology, and in particular to a method, device and electronic equipment for calculating the position of a resource management product. Background Art

[0002] A public fund refers to a resource management product that raises resources from the public in an open manner, invests the raised resources into investment products in specific industries, and obtains returns for the invested resources based on changes in industry returns.

[0003] In order to allow the subject to select the appropriate mutual fund products, it is necessary to properly disclose the investment information of mutual funds. Among them, the industry position information of mutual funds is an important information of mutual funds, and the fund management institution needs to calculate and disclose the industry position of mutual funds in a timely manner.

[0004] The current methods used in related technologies to calculate industry positions for resource management products such as public funds have low accuracy in calculating industry positions. Summary of the invention

[0005] The embodiments of the present disclosure provide a method, device and electronic device for calculating the position of a resource management product. The method for calculating the position of a resource management product can improve the accuracy of calculating the industry position of the resource management product.

[0006] A first aspect of the present disclosure provides a method for calculating the position of a resource management product, the method comprising:

[0007] Obtain an industry position calculation request for a target resource management product, wherein the industry position calculation request includes target industry information and calculation time information;

[0008] Acquire first historical revenue statistics of the target resource management product and acquire second historical revenue statistics corresponding to the target industry information;

[0009] Determine a plurality of measurement data collection time points based on the measurement time information, and determine a revenue relationship data sequence corresponding to the measurement time information according to the plurality of measurement data collection time points, the first historical revenue statistical data, and the second historical revenue statistical data;

[0010] The target neural network model is called to perform position calculation on the revenue relationship data sequence to obtain the position information corresponding to the target industry in the target resource management product corresponding to the calculation time information.

[0011] A second aspect of the present disclosure provides a device for calculating the position of a resource management product, the device comprising:

[0012] A first acquisition unit is used to acquire an industry position calculation request of a target resource management product, wherein the industry position calculation request includes target industry information and calculation time information;

[0013] A second acquisition unit is used to acquire first historical revenue statistical data of the target resource management product and to acquire second historical revenue statistical data corresponding to the target industry information;

[0014] a determining unit, configured to determine a plurality of measurement data collection time points based on the measurement time information, and determine a revenue relationship data sequence corresponding to the measurement time information according to the plurality of measurement data collection time points, the first historical revenue statistical data, and the second historical revenue statistical data;

[0015] The calculation unit is used to call the target neural network model to perform position calculation on the revenue relationship data sequence, and obtain the position information corresponding to the target industry in the target resource management product corresponding to the calculation time information.

[0016] Optionally, in one implementation, the determining unit is specifically configured to:

[0017] Extracting a plurality of node yield data from the first historical yield statistical data based on the plurality of measured data collection time points, and extracting a plurality of industry yield data from the second historical yield statistical data;

[0018] The profit relationship data corresponding to each of the measurement data collection time points is determined according to the node yield data and the corresponding industry yield data, and the profit relationship data is sorted according to the time sequence of the multiple measurement data collection time points to obtain a profit relationship data sequence corresponding to the measurement time information.

[0019] Optionally, in one implementation, the determining unit is specifically configured to:

[0020] Calculate the difference between the node yield data and the industry yield data corresponding to each of the measured data collection time points to obtain the first sub-yield relationship data;

[0021] Calculate the correlation coefficient between the first yield trend corresponding to the node yield data corresponding to each of the measured data collection time points and the second yield trend corresponding to the industry yield data to obtain the second sub-yield relationship data;

[0022] Perform correlation fitting between the node yield data and the industry yield data corresponding to each of the measured data collection time points, and determine the fitting parameters obtained by fitting as the third sub-yield relationship data;

[0023] The revenue relationship data corresponding to each of the measurement data collection time points is determined according to the first sub-revenue relationship data, the second sub-revenue relationship data and the third sub-revenue relationship data.

[0024] Optionally, in one implementation, the determining unit is specifically configured to:

[0025] Obtain product yield values ​​and industry yield values ​​in multiple first nearest neighbor time periods corresponding to each of the measured data collection time points;

[0026] Calculate the difference between the product yield value and the industry yield value corresponding to each of the first neighboring time periods to obtain multiple yield difference values ​​corresponding to the multiple first neighboring time periods, and determine the first sub-yield relationship data based on the multiple yield difference values.

[0027] Optionally, in one implementation, the determining unit is specifically configured to:

[0028] Obtain product daily return trend data and industry daily return trend data in multiple second nearest neighbor time periods corresponding to each of the measured data collection time points;

[0029] Calculate the correlation coefficient between the product daily return trend data and the industry daily return trend data corresponding to each of the second neighboring time periods, obtain multiple correlation coefficients corresponding to the multiple second neighboring time periods, and determine the second sub-return relationship data based on the multiple correlation coefficients.

[0030] Optionally, in one implementation, the determining unit is specifically configured to:

[0031] Obtain product daily yield sequences and industry yield sequences in multiple third nearest neighbor time periods corresponding to each of the measured data collection time points;

[0032] Performing linear fitting on the product daily yield sequence and the industry yield sequence to obtain a first linear fitting result;

[0033] Determine a first fitting parameter corresponding to each of the third nearest neighbor time periods based on a slope coefficient of the first linear fitting result, and determine a second fitting parameter corresponding to each of the third nearest neighbor time periods based on a residual square sum of the first linear fitting result;

[0034] The third sub-income relationship data is determined according to the first fitting parameter and the second fitting parameter.

[0035] Optionally, in one implementation, the training process of the target neural network model includes:

[0036] Acquire multiple training sample data, wherein the training sample data includes a sample income relationship data sequence between a sample resource management product and a sample industry and a corresponding industry position label sequence, wherein the sample income relationship data sequence is a sequence of sample income relationship data corresponding to multiple sample time nodes in chronological order;

[0037] The sample return relationship data sequence in the multiple training sample data is used as the model input, and the corresponding industry position label sequence is used as the model training label training preset recurrent neural network model to obtain the trained target neural network model.

[0038] Optionally, in one implementation, a plurality of training sample data are obtained, the training sample data comprising a sample revenue relationship data sequence between a sample resource management product and a sample industry and a corresponding industry position label sequence, the sample revenue relationship data sequence being a sequence of sample revenue relationship data corresponding to a plurality of sample time nodes in chronological order, including:

[0039] Obtain third historical revenue statistics and historical industry position information statistics of multiple sample resource management products within a preset period of time, and obtain industry revenue statistics corresponding to resource management products of multiple industries within the preset period of time;

[0040] Determining a plurality of sample time nodes within the preset period, and determining a sample industry among the plurality of industries;

[0041] Extracting a sample revenue relationship data sequence corresponding to each of the sample resource management products from the third historical revenue statistical data and the industry revenue statistical data based on the multiple sample time nodes and the sample industry;

[0042] Determine the industry position label sequence corresponding to each of the sample return relationship data sequences based on the multiple sample time nodes and the historical industry position information statistical data;

[0043] A plurality of training sample data are determined based on the sample revenue relationship data sequences of the plurality of sample resource management products and sample industries and the corresponding industry position label sequences.

[0044] Optionally, in one implementation, before determining a plurality of training sample data according to the plurality of sample resource management products and sample industry sample revenue relationship data sequences and corresponding industry position label sequences, the method further includes:

[0045] Performing linear interpolation processing on the sample income relationship data sequence and the corresponding industry position label sequence corresponding to each of the sample resource management products to obtain first sub-enhanced sample data;

[0046] Performing data splicing on the sample income relationship data sequences and the corresponding industry position label sequences corresponding to the multiple sample resource management products to obtain second sub-enhanced sample data;

[0047] Determine enhanced sample data according to the first sub-enhanced sample data and the second sub-enhanced sample data;

[0048] The determining of a plurality of training sample data according to the plurality of sample resource management products and sample industry sample revenue relationship data sequences and the corresponding industry position label sequences comprises:

[0049] A plurality of training sample data are determined according to the enhanced sample data, the sample revenue relationship data sequences of the plurality of sample resource management products and sample industries, and the corresponding industry position label sequences.

[0050] Optionally, in one implementation, performing linear interpolation processing on the sample revenue relationship data sequence and the corresponding industry position label sequence corresponding to each of the sample resource management products to obtain the first sub-enhanced sample data includes:

[0051] Obtaining the first sampling time and the first industry position information corresponding to each sample benefit relationship data in the sample benefit relationship data sequence corresponding to each of the sample resource management products;

[0052] Performing a linear fit on the first sampling time and the first industry position information to obtain a second linear fit result;

[0053] Perform interpolation processing based on the second linear fitting result to obtain multiple second sampling times and corresponding second industry position information;

[0054] Obtain enhanced revenue relationship data corresponding to each of the second sampling times, and determine first sub-enhanced sample data based on the enhanced revenue relationship data and the second industry position information.

[0055] Optionally, in one implementation, the data splicing of the sample revenue relationship data sequences and the corresponding industry position label sequences corresponding to the multiple sample resource management products to obtain the second sub-enhanced sample data includes:

[0056] Perform feature concatenation on the sample revenue relationship data sequences corresponding to multiple sample resource management products to obtain enhanced sample features;

[0057] The industry position label sequences corresponding to the multiple sample resource management products are sequentially spliced ​​to obtain enhanced sample labels, and second sub-enhanced sample data is determined according to the enhanced sample features and the enhanced sample labels.

[0058] Optionally, in one embodiment, the sample return relationship data sequence in the multiple training sample data is used as a model input, and the corresponding industry position label sequence is used as a model training label training preset recurrent neural network model to obtain a trained target neural network model, including:

[0059] Dividing the plurality of training sample data into a first training set and a first validation set;

[0060] Randomly extract a preset proportion of target training sample data from the first training set, and construct a second validation set with the extracted target training sample data and the first validation set, and construct a second training set with the training sample data other than the target training sample data in the first training set;

[0061] A preset long short-term memory network model is trained based on the sample return relationship data sequence in the second training set as a model input and the corresponding industry position label sequence as a model training label;

[0062] The trained long short-term memory network model is verified according to the second verification set, and when the verification result meets the preset verification condition, the trained long short-term memory network model is determined as the target neural network model.

[0063] A third aspect of the present disclosure provides a storage medium storing a computer program, which, when executed by a processor, implements the method for calculating the position of a resource management product as described in the first aspect.

[0064] A fourth aspect of the present disclosure provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the method for calculating the position of a resource management product as described in the first aspect is implemented.

[0065] A fifth aspect of the present disclosure provides a computer program product, which includes a computer program, and the computer program is read and executed by a processor of an electronic device, so that the electronic device executes the position calculation method of the resource management product as described in the first aspect.

[0066] The presently disclosed embodiment provides a method for calculating the position of a resource management product, which obtains an industry position calculation request for a target resource management product, wherein the industry position calculation request includes target industry information and calculation time information; obtains first historical revenue statistical data of the target resource management product and obtains second historical revenue statistical data corresponding to the target industry information; determines a plurality of calculation data collection time points based on the calculation time information, and determines a revenue relationship data sequence corresponding to the calculation time information based on the plurality of calculation data collection time points, the first historical revenue statistical data, and the second historical revenue statistical data; calls a target neural network model to perform position calculation on the revenue relationship data sequence, and obtains the position information corresponding to the target industry in the target resource management product corresponding to the calculation time information.

[0067] In this way, the embodiment of the present disclosure determines the revenue relationship data sequence based on multiple measurement data collection time points corresponding to the measurement time information, the first historical revenue statistical data corresponding to the target resource management product, and the second historical revenue statistical data corresponding to the target industry. Through the revenue relationship data sequence, the target neural network model can analyze the first historical revenue statistical data and the second historical revenue statistical data corresponding to multiple measurement data collection time points to predict the position information corresponding to the target industry in the target resource management product at the measurement time. Therefore, in the position measurement method of the resource management product of the embodiment of the present disclosure, the target neural network model can use the revenue-related information of the target resource management product and the target industry collected at multiple measurement data collection time points in the revenue relationship data sequence, and the long-term dependence of the revenue-related information in time series as a basis to predict the position information corresponding to the target industry at the measurement time, thereby improving the accuracy of measuring the industry position corresponding to the resource management product.

[0068] Other features and advantages of the present disclosure will be described in the following description, and partly become apparent from the description, or understood by practicing the present disclosure. The purpose and other advantages of the present disclosure can be realized and obtained by the structures particularly pointed out in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] The accompanying drawings are used to provide further understanding of the technical solution of the present disclosure and constitute a part of the specification. Together with the embodiments of the present disclosure, they are used to explain the technical solution of the present disclosure and do not constitute a limitation on the technical solution of the present disclosure.

[0070] Figure 1 A system architecture diagram for a method for calculating the position of a resource management product according to an embodiment of the present disclosure;

[0071] Figure 2 It is a schematic diagram of an embodiment of the present disclosure applied to a scenario of evaluating industry conditions based on public funds;

[0072] Figure 3 It is a schematic diagram of an embodiment of the present disclosure applied to a scenario of constructing a public fund industry portrait;

[0073] Figure 4 It is a schematic diagram of an embodiment of the present disclosure applied to a scenario of building a public fund recommendation system based on an industry;

[0074] Figure 5 is a flow chart of a method for calculating a position of a resource management product according to an embodiment of the present disclosure;

[0075] Figure 6 is a schematic diagram of resource exchange performed by a resource management product according to an embodiment of the present disclosure;

[0076] Figure 7 It is a schematic diagram of the yield change trend of a target resource management product within one month according to an embodiment of the present disclosure;

[0077] Figure 8 It is a schematic diagram of the yield change trend of the target industry within one month according to an embodiment of the present disclosure;

[0078] Fig. 9 is a schematic diagram of a first linear fitting result between a product yield and an industry yield in a third nearest neighbor period according to an embodiment of the present disclosure;

[0079] Fig.10 is a schematic diagram of sorting revenue relationship data in chronological order to form a revenue relationship data sequence according to an embodiment of the present disclosure;

[0080] Fig.11 is a schematic diagram of obtaining a second linear fitting result according to an embodiment of the present disclosure;

[0081] Fig.12 is a schematic diagram of adding connection variables to sample revenue relationship data according to an embodiment of the present disclosure;

[0082] Fig.13 It is a schematic diagram of feature sensitivity when a preset recurrent neural network model according to an embodiment of the present disclosure is learned based on enhanced sample features obtained by data splicing after adding connection variables;

[0083] Fig.14 is a schematic diagram of data processing by a neuron in a long short-term memory network according to an embodiment of the present disclosure;

[0084] Fig.15is a schematic diagram of processing a sample return relationship data sequence based on a long short-term memory network according to an embodiment of the present disclosure;

[0085] Fig.16 is a schematic diagram of a preset recurrent neural network structure according to an embodiment of the present disclosure;

[0086] Fig.17 is a schematic diagram of dividing a training sample data set according to an embodiment of the present disclosure;

[0087] Fig.18 is a flow chart of a training process of a target neural network model according to an embodiment of the present disclosure;

[0088] Fig.19 Another flowchart of the method for calculating the position of the resource management product provided by the present disclosure;

[0089] Fig. 20 A processing schematic diagram of a method for calculating a position of a resource management product according to an embodiment of the present disclosure;

[0090] Fig.21 A schematic diagram of the structure of a storage location calculation device for a resource management product provided in an embodiment of the present disclosure;

[0091] Fig. 22 is a terminal structure diagram for implementing various methods according to an embodiment of the present disclosure;

[0092] Fig.23 It is a server structure diagram for implementing various methods according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0093] In order to make the purpose, technical solution and advantages of the present disclosure more clear, the present disclosure is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present disclosure and are not used to limit the present disclosure.

[0094] Before further describing the embodiments of the present disclosure in detail, the nouns and terms involved in the embodiments of the present disclosure are described. The nouns and terms involved in the embodiments of the present disclosure are subject to the following interpretations:

[0095] Recurrent Neural Network (RNN): A neural network model that can process sequence data. RNN retains a hidden state when processing each input, which is passed to the next time step so that the model can remember previous information.

[0096] Long Short-Term Memory (LSTM): A special type of recurrent neural network (RNN) designed specifically to solve the problem of long-term dependencies, that is, to remember information over longer time intervals when processing sequence data.

[0097] LSTM has three gates: forget gate, input gate and output gate, as well as a cell state and hidden state. The forget gate determines which old information is forgotten, the input gate controls the acceptance of new information, and the output gate determines how much information in the cell state will be used to generate the hidden state.

[0098] The Pearson Correlation Coefficient is a statistic that measures the degree of linear correlation between two variables, and its value is between -1 and 1. When the correlation coefficient is greater than 0, it means that the two variables are positively correlated; when the correlation coefficient is less than 0, it means that the two variables are negatively correlated; when the correlation coefficient is equal to 0, it means that there is no linear relationship between the two variables. The Pearson correlation coefficient is defined as the ratio of the covariance between two variables X and Y and the product of their respective standard deviations.

[0099] Sigmoid activation function: is a nonlinear function, usually used in the hidden layer or output layer of a neural network to introduce nonlinear characteristics. It maps the input value to an output ranging between 0 and 1.

[0100] Tanh layer: commonly used in hidden layers of neural networks because it can produce outputs in a range centered on zero, which helps convergence during training. In LSTM networks, the Tanh layer is used to calculate updates to the cell state and hidden state, helping the network learn long-term dependencies.

[0101] Mean Absolute Error (MAE): A commonly used regression loss function. It calculates the average of the absolute values ​​of the differences between the predicted values ​​and the true values.

[0102] In the related art, the industry position data of resource management products within a predetermined time period will be published according to a predetermined time cycle. When calculating the industry position of resource management products at a specific time node, if the industry position data is not published at that time node, the most recent industry position data can be used for analysis. The data calculated in this way has a lag and cannot accurately represent the industry position information of the resource management product at that time node.

[0103] In order to solve the timeliness problem of industry position data, algorithms such as linear regression or Kalman filtering can be used to construct a linear relationship between resource management product revenue data, industry revenue data, and industry position information, and calculate the industry position of resource management products at specific time points based on the linear relationship. However, resource management product revenue data and industry revenue data are characterized by high correlation and low signal-to-noise ratio, and revenue data has a long-term dependency in time series, so it cannot be described by a linear relationship. Therefore, the accuracy of calculating industry position information in related technologies is low. In this regard, the present application provides a position calculation method for a resource management product, in order to improve the accuracy of calculating the industry position information corresponding to the resource management product.

[0104] System architecture and scenario description of the application of the embodiments of the present disclosure

[0105] Figure 1 It is a system architecture diagram applied to the position calculation method of the resource management product according to the embodiment of the present disclosure, and includes a terminal 140, the Internet 130, a gateway 120, a server 110, and the like.

[0106] The terminal 140 includes a variety of device forms with a display screen, such as a desktop computer, a laptop computer, a PDA (personal digital assistant), a mobile phone, a vehicle terminal, a home theater terminal, a dedicated terminal, an intelligent voice interaction device, an intelligent home appliance or an aircraft. In addition, it can be a single device or a collection of multiple devices. The terminal 140 can communicate with the Internet 130 in a wired or wireless manner to exchange data.

[0107] The server 110 refers to a computer system that can provide certain services to the terminal 140. Compared with the ordinary terminal 140, the server 110 has higher requirements in terms of stability, security, performance, etc. The server 110 can be a high-performance computer in a network platform, a cluster of multiple high-performance computers, a part of a high-performance computer (such as a virtual machine), a combination of parts of multiple high-performance computers (such as virtual machines), etc.

[0108] The gateway 120 is also called an internetwork connector or a protocol converter. The gateway realizes network interconnection at the transport layer and is a computer system or device that acts as a converter. The gateway is a translator between two systems that use different communication protocols, data formats or languages, or even completely different architectures. At the same time, the gateway can also provide filtering and security functions. The message sent by the terminal 140 to the server 110 must be sent to the corresponding server 110 through the gateway 120. The message sent by the server 110 to the terminal 140 must also be sent to the corresponding terminal 140 through the gateway 120.

[0109] The position calculation method of the resource management product provided in the embodiment of the present disclosure may be implemented solely in the terminal 140 , may be implemented solely in the server 110 , or may be implemented partially in the terminal 140 and partially in the server 110 .

[0110] When the position calculation method of the resource management product provided by the embodiment of the present disclosure is implemented separately in the terminal 140, the terminal 140 obtains the industry position calculation request of the target resource management product; the terminal 140 obtains the first historical revenue statistical data of the target resource management product and obtains the second historical revenue statistical data corresponding to the target industry information; the terminal 140 determines a plurality of calculation data collection time points based on the calculation time information, and determines the revenue relationship data sequence corresponding to the calculation time information according to the plurality of calculation data collection time points, the first historical revenue statistical data and the second historical revenue statistical data; the terminal 140 calls the target neural network model to perform position calculation on the target revenue relationship data sequence, and obtains the position information corresponding to the target industry in the target resource management product at the target calculation moment.

[0111] When the position calculation method of the resource management product provided by the embodiment of the present disclosure is implemented separately in the server 110, the server 110 obtains the industry position calculation request of the target resource management product; the server 110 obtains the first historical revenue statistical data of the target resource management product and obtains the second historical revenue statistical data corresponding to the target industry information; the server 110 determines a plurality of measurement data collection time points based on the measurement time information, and determines the revenue relationship data sequence corresponding to the measurement time information according to the plurality of measurement data collection time points, the first historical revenue statistical data, and the second historical revenue statistical data; the server 110 calls the target neural network model to perform position calculation on the target revenue relationship data sequence, and obtains the position information corresponding to the target industry in the target resource management product at the target measurement moment.

[0112] When the position calculation method of the resource management product provided by the embodiment of the present disclosure is partially implemented in the terminal 140 and partially implemented in the server 110, the terminal 140 obtains the industry position calculation request of the target resource management product; the server 110 calls the industry position calculation request of the terminal 140, obtains the first historical revenue statistical data of the target resource management product and obtains the second historical revenue statistical data corresponding to the target industry information; the server 110 determines a plurality of calculation data collection time points based on the calculation time information, and determines the revenue relationship data sequence corresponding to the calculation time information according to the plurality of calculation data collection time points, the first historical revenue statistical data and the second historical revenue statistical data; the server 110 calls the target neural network model to perform position calculation on the target revenue relationship data sequence, and obtains the position information corresponding to the target industry in the target resource management product at the target calculation moment.

[0113] The embodiments of the present disclosure can be applied in various scenarios, such as Figure 2 The scenario shown is based on the assessment of industry conditions by public funds; for example Figure 3 The scenario of building a public fund industry portrait is shown in the figure; for example Figure 4 The scenario shown is of building a public fund recommendation system based on the industry.

[0114] Public funds are a type of resource management product. Since the development status of each industry can determine the return status of the public funds invested in that industry, the development status of the industry can be judged by calculating the industry positions of public funds and then based on the development trends of multiple public funds corresponding to each industry. Figure 2 As shown, the industry status of industry I1, industry I2, industry I3 and industry I4 can be viewed. Among them, the public funds that occupy a large position in industry I1 are determined by the position calculation method of the resource management product of the embodiment of the present disclosure, including public funds A1 and public funds A2. By comprehensively analyzing the development trends of public funds A1 and public funds A2, the development status of industry I1 can be obtained. Its short-term development is good, which can reach a score of 4.2, and its long-term development is also good, which can reach a score of 4.3.

[0115] When the subject understands each public fund, it is very important to understand the industry composition of the public fund and the positions occupied by each industry. Therefore, it is very important to obtain accurate industry position information using the position calculation method of the resource management product of the embodiment of the present disclosure. Figure 3 As shown, the object can first see the industry with the highest position in public fund A1, that is, industry I1. Further, the object can also view more detailed industry information of public fund A1, including: industry I1, industry I2, industry I3, industry I4, industry I5 and industry I6, where the display size corresponding to different industries can reflect the position size of the industry in public fund A1. This example does not limit the display method of industry position information.

[0116] When the subject determines the industry with a good development trend, the public fund can be selected based on the industry. The position calculation method of the resource management product of the embodiment of the present disclosure can be used to determine the industry composition of each public fund and the position occupied by each industry. Therefore, a public fund recommendation system can be constructed based on the industry composition and industry position information. Figure 4As shown in the figure, the industry sector includes industry I1 and industry I2, among which the public fund with a larger position in industry I1 is public fund A1, and the public fund with a larger position in industry I2 is public fund A2. Taking industry I1 as an example, based on industry I1, more public fund information can be obtained, including public fund A3, public fund A4 and public fund A2 in addition to public fund A1. Different resource management products can be sorted according to the size of the industry position of the corresponding industry, so that the object can choose a public fund based on the target industry.

[0117] The above examples do not limit the scope of protection of this case.

[0118] General description of the disclosed embodiments

[0119] According to an embodiment of the present disclosure, a method for calculating the position of a resource management product is provided. Figure 5 FIG. 1 is a flow chart of a method for calculating the position of a resource management product provided by the present disclosure. The method can be applied to a device for calculating the position of a resource management product, and the device for calculating the position of a resource management product can be integrated in an electronic device, which can be a terminal or a server. The method for calculating the position of a resource management product can include:

[0120] Step 510: Obtain an industry position calculation request for a target resource management product.

[0121] The target resource management product may be a resource management product that requires industry position information measurement. Resource management products may be products that collect resources from the target in exchange for resources in a specific industry to obtain benefits for the target, such as public funds. Figure 6 As shown, after objects A1, A2, and A3 invest resources into the resource management product, the resource management product can exchange resources from industries I1, I2, and I3 according to the predetermined industry position information based on the amount of resources invested by each object. When the value of resources in industries I1, I2, and I3 changes, the value of resources held by each object will also change, so that the object can obtain benefits through the resource management product.

[0122] The industry position calculation request includes target industry information and calculation time information. The target industry information may indicate the industry information whose position in the target resource management product needs to be calculated. The calculation time information may indicate the time for calculating the industry position information of the target resource management product. The calculation time may be the time when the industry position calculation request is obtained or a specified time.

[0123] The industry position calculation request may be generated by an object that wants to obtain the industry position information of the target resource management product; or it may be generated according to a predetermined calculation time period.

[0124] Step 520: Obtain first historical revenue statistics of the target resource management product and obtain second historical revenue statistics corresponding to the target industry information.

[0125] The first historical revenue statistical data may be revenue-related data of the target resource management product at different times within a preset period, such as the yield data of the target resource management product within the preset period, etc. For example, at time T1 within the preset period, the yield of the target resource management product is 8%; at time T2 within the preset period, the yield of the target resource management product is 10%.

[0126] In one embodiment, the first historical revenue statistical data is statistically recorded in the form of a resource management product revenue report. Therefore, the first historical revenue statistical data of the target resource management product can be obtained by obtaining the resource management product revenue report of the target resource management product within a preset time period. The resource management product revenue report can be a regularly counted report used to show the target resource management product's open and transparent revenue situation to the object. The resource management product revenue report can have multiple types depending on the statistical cycle, for example, a monthly report (statistical disclosure with a monthly cycle), a quarterly report (statistical disclosure with a quarterly cycle), a semi-annual report (statistical disclosure with a half-year cycle), and an annual report (statistical disclosure with a year as a cycle). The resource management product revenue report may include the revenue situation of the target resource management product over a period of time. According to different preset time periods, the resource management product revenue report corresponding to the target resource management product, whose statistical time includes the predetermined time period, can be obtained, and then the first historical revenue statistical data can be extracted from the resource management product revenue report.

[0127] The second historical revenue statistical data may be the revenue of the target industry within a preset period of time. The revenue of the target industry may be determined by the revenue of the resource management products invested in the target industry within a preset period of time. When the industry revenue is high, the resource management products that include the industry will also obtain higher revenue in the industry; however, when the industry revenue is low or even incurring losses, the resource management products that include the industry will not be able to obtain higher revenue in the industry. The higher the position of the industry in the resource management product, the closer the revenue of the resource-related product will be to the industry revenue. If the position of the industry in the resource management product is high, the revenue of the industry may directly affect the overall revenue of the resource management product.

[0128] In one implementation, obtaining second historical revenue statistical data corresponding to target industry information includes:

[0129] Obtain industry revenue statistics of multiple resource management products corresponding to the target industry in the target industry;

[0130] The second historical revenue statistical data is determined based on the industry revenue statistical data corresponding to the plurality of resource management products.

[0131] The multiple resource management products corresponding to the target industry may be resource management products that include the target industry. The industry revenue statistics of the resource management products in the target industry may also be obtained through the resource management product revenue report. The resource management product revenue report may not only count the revenue data of the corresponding resource management products, but may also include industry position information data of multiple industries and independent industry revenue statistics of each industry. Therefore, the industry revenue statistics corresponding to the target industry may be extracted from the resource management product revenue report corresponding to the multiple resource management products that include the target industry.

[0132] After obtaining the industry revenue statistics, the second historical revenue statistics can be determined based on the industry revenue statistics corresponding to the multiple resource management products. Specifically, the industry revenue statistics corresponding to the multiple resource management products can be summarized and counted according to a predetermined statistical rule to obtain the second historical revenue statistics.

[0133] Step 530: determine multiple measurement data collection time points based on the measurement time information, and determine the revenue relationship data sequence corresponding to the measurement time information according to the multiple measurement data collection time points, the first historical revenue statistical data, and the second historical revenue statistical data.

[0134] The time point for collecting the measured data may be a time point for extracting the industry position measurement related data from the first historical income statistical data and the second historical income statistical data. The multiple time points for collecting the measured data may be multiple random time points within a preset period before the measurement time, or may be multiple time points determined according to a preset cycle. For example, if the preset period is one month before the measurement time, the time point for collecting the measured data may be determined in a weekly preset cycle. Within the preset period, a time point may be determined as the time point for collecting the measured data every one week. In this way, four time points for collecting the measured data may be determined within the preset period.

[0135] After determining a plurality of measurement data collection time points, the revenue relationship data sequence corresponding to the measurement time information may be determined based on the plurality of measurement data collection time points, the first historical revenue statistical data, and the second historical revenue statistical data.

[0136] The revenue relationship data sequence may be a sequence of revenue relationship data corresponding to multiple measurement data collection time points sorted in chronological order. The revenue relationship data may be obtained from the first historical revenue statistical data and the second historical revenue statistical data based on the corresponding measurement data collection time points.

[0137] In one implementation, determining the revenue relationship data sequence corresponding to each measurement time information according to multiple measurement data collection time points, the first historical revenue statistical data, and the second historical revenue statistical data includes:

[0138] Extracting multiple node yield data from the first historical yield statistical data based on multiple measurement data collection time points, and extracting multiple industry yield data from the second historical yield statistical data;

[0139] The income relationship data corresponding to each measurement data collection time point is determined according to the node yield data and the corresponding industry yield data, and the income relationship data is sorted according to the time sequence of multiple measurement data collection time points to obtain the income relationship data sequence corresponding to the measurement time information.

[0140] First, multiple node yield data corresponding to multiple measurement data collection time points can be extracted from the first historical yield statistical data. The node yield data can represent the yield data of the target resource management product at the corresponding measurement data collection time point. For example Figure 7 What is shown is the yield change trend of the target resource management product within one month. The multiple measurement data collection time points determined during this period are T1, T2, T3 and T4. Through the yield change trend, it can be determined that the node yield data corresponding to T1 is W1, the node yield data corresponding to T2 is W2, the node yield data corresponding to T3 is W3, and the node yield data corresponding to T4 is W4.

[0141] Extract multiple industry yield data corresponding to the target industry at multiple measurement data collection time points from the second historical yield statistical data. The industry yield data may represent the yield data of the target industry at the corresponding measurement data collection time points. For example Figure 8 The yield change trend of the target industry within one month is shown. From the yield change trend, it can be determined that the industry yield data corresponding to T1 is Q1, the industry yield data corresponding to T2 is Q2, the industry yield data corresponding to T3 is Q3, and the industry yield data corresponding to T4 is Q4.

[0142] After extracting multiple node yield data corresponding to the target resource management product at multiple measurement data collection time points and multiple industry yield data corresponding to the target industry at multiple measurement data collection time points, the yield relationship data corresponding to each measurement data collection time point can be calculated based on the node yield data and the corresponding industry yield data. The yield relationship data can reflect the relationship between the target resource product yield and the target industry yield at the corresponding measurement data collection time point.

[0143] In one implementation, the yield relationship data corresponding to each measurement data collection time point is determined based on the node yield data and the corresponding industry yield data, including:

[0144] Calculate the difference between the node yield data and the industry yield data corresponding to each measurement data collection time point to obtain the first sub-yield relationship data;

[0145] Calculate the correlation coefficient between the first yield trend corresponding to the node yield data corresponding to each measured data collection time point and the second yield trend corresponding to the industry yield data to obtain the second sub-yield relationship data;

[0146] Perform correlation fitting between the node yield data and the industry yield data corresponding to each measurement data collection time point, and determine the fitting parameters obtained by fitting as the third sub-yield relationship data;

[0147] The revenue relationship data corresponding to each measurement data collection time point is determined according to the first sub-revenue relationship data, the second sub-revenue relationship data and the third sub-revenue relationship data.

[0148] The first sub-income relationship data can represent the difference between the node yield data corresponding to the measurement data collection time point and the corresponding industry yield data. Since the higher the position of the target industry in the target resource management product, the closer the target resource management product's yield trend is to the target industry's yield trend. Therefore, the first sub-income relationship data can be used to describe the similarity of the returns between the target resource management product and the target industry. The smaller the first sub-income relationship data is, the smaller the yield difference between the target resource management product and the target industry at the corresponding measurement data collection time point is, the higher the similarity is, and the higher the position of the target industry in the target resource management product is.

[0149] In one implementation, the difference between the node yield data and the industry yield data corresponding to each measurement data collection time point is calculated to obtain the first sub-yield relationship data, including:

[0150] Obtain product yield values ​​and industry yield values ​​in multiple first nearest neighbor time periods corresponding to each measured data collection time point;

[0151] Calculate the difference between the product yield value and the industry yield value corresponding to each first neighbor time period, obtain multiple yield difference values ​​corresponding to multiple first neighbor time periods, and determine the first sub-yield relationship data based on the multiple yield difference values.

[0152] The measurement data collection time point is a time point within a preset time period. In this embodiment, a plurality of yield differences of a plurality of first adjacent time periods corresponding to the measurement data collection time point may be obtained to determine the first sub-yield relationship data.

[0153] Since the yield data is constantly changing, the yield data at a single time point may have outliers within a period of time, which is inconsistent with the overall yield change during this period. Therefore, when determining the difference between the node yield data corresponding to the time point of the measured data collection and the industry yield data, the yield difference of the first nearest neighbor period corresponding to the time point of the measured data collection can be obtained. Therefore, using the yield difference in multiple first nearest neighbor periods corresponding to the time point of the measured data collection can improve the accuracy of determining the similarity between the target resource management product and the target industry.

[0154] The first neighboring time period may be a preset neighboring time period before the time point of the measurement data collection, for example, one day or one week before the time point of the measurement data collection.

[0155] The product yield value in the first neighboring time period can be obtained based on the first historical yield statistical data. The first historical yield statistical data can record the yield data of multiple time points in the preset time period. Based on the starting net income value and the ending net income value in the first neighboring time period in the first historical yield statistical data, the product yield value of the first neighboring time period can be calculated. When the first neighboring time period is the week before the measurement data collection time point, the product yield value can be the result of deducting the starting net income value from the ending net income value of the week and dividing it by the starting net income value. For example, if the ending net income value of this week is 1.52 and the starting net income value is 1.5, then the product yield value of this week is equal to (1.52-1.5) / 1.5*100%=1.33%.

[0156] The industry yield value can be obtained based on the second historical yield statistical data. The method of determining the industry yield value corresponding to the first neighboring time period based on the second historical yield statistical data is similar to the method of calculating the product yield value based on the first historical yield statistical data, which will not be repeated here.

[0157] After obtaining the product yield values ​​and industry yield values ​​in multiple first neighbor time periods, the difference between the product yield value and the industry yield value corresponding to each first neighbor time period can be calculated to obtain the difference between the target resource management product and the target industry in the first neighbor time period. For example, when the first neighbor time period is one day before the measurement data collection time point, the yield difference is the difference between the daily yield value of the target resource management product and the daily yield value of the target industry; when the first neighbor time period is one week before the measurement data collection time point, the yield difference is the difference between the weekly yield value of the target resource management product and the weekly yield value of the target industry.

[0158] The first sub-yield relationship data may include multiple yield differences corresponding to multiple first neighboring time periods.

[0159] The second sub-income relationship data can be used to describe the correlation between the first income trend corresponding to the node income data and the second income trend corresponding to the industry income data. The first income trend describes the income change of the target resource management product at the time of measurement data collection; the second income trend describes the income change of the target industry at the time of measurement data collection. The higher the correlation between the income trend of the target resource management product and the target industry at the time of measurement data collection, the greater the impact of the income change of the target industry on the income change of the target resource management product, and the larger the position of the target industry in the target resource management product.

[0160] The correlation coefficient may be expressed as a Pearson correlation coefficient. Therefore, the correlation coefficient between the first return trend and the second return trend may be equal to the ratio of the covariance between the first return trend and the second return trend and the product of their respective standard deviations.

[0161] In one implementation, the correlation coefficient between the first yield trend corresponding to the node yield data corresponding to each measurement data collection time point and the second yield trend corresponding to the industry yield data is calculated to obtain the second sub-yield relationship data, including:

[0162] Obtain product daily return trend data and industry daily return trend data in multiple second nearest neighbor time periods corresponding to each measured data collection time point;

[0163] Calculate the correlation coefficient between the product daily return trend data and the industry daily return trend data corresponding to each second nearest neighbor time period, obtain multiple correlation coefficients corresponding to multiple second nearest neighbor time periods, and determine the second sub-return relationship data based on the multiple correlation coefficients.

[0164] The second nearest neighbor period may be a nearest neighbor period before the measurement data collection time point, for example, nearly one month, nearly two months, nearly three months, nearly six months, and nearly one year before the measurement data collection time point.

[0165] Since the return trend is used to describe the change in return, the first return trend corresponding to the return rate data corresponding to the time point of the measured data collection and the second return trend corresponding to the industry return rate data can only describe the return change at a very short time point. The product daily return trend data in the second nearest neighbor period can describe the daily return change of the target resource management product in the second nearest neighbor period, and the industry daily return trend data can describe the daily return change of the target industry in the second nearest neighbor period. Determining the second sub-return relationship data corresponding to the time point of the measured data collection by using the correlation coefficient of the long-term return changes of the target resource management product and the target industry can improve the accuracy of obtaining the correlation between the return trend of the target resource management product and the return trend of the target industry.

[0166] The product daily revenue trend data in multiple second neighboring time periods can be obtained based on the first historical revenue statistical data. The first historical revenue statistical data can record the revenue data of multiple time points in a preset time period. The product daily revenue trend data in the second neighboring time period is generated by obtaining the daily revenue data change of the target resource management product in the second neighboring time period from the first historical revenue statistical data.

[0167] The industry daily income trend data can be obtained based on the second historical income statistical data. The method of determining the industry daily income trend data corresponding to the second nearest time period based on the second historical income statistical data is similar to the method of obtaining the product daily income trend data based on the first historical income statistical data, which will not be repeated here.

[0168] After obtaining the product daily return trend data and the industry daily return trend data in multiple second nearest neighbor time periods, the correlation coefficient between the product daily return trend data and the industry daily return trend data corresponding to each second nearest neighbor time period can be calculated. Specifically, the correlation coefficient can be obtained by calculating the Pearson correlation coefficient between the product daily return trend data and the industry daily return trend data.

[0169] The second sub-income relationship data corresponding to the measured data collection time point may include multiple correlation coefficients corresponding to multiple second neighboring time periods.

[0170] The third sub-income relationship data can be used to describe the relative change relationship between the node yield data and the industry yield data corresponding to the measurement data collection time point. Correlation fitting can fit the change correlation between the node yield data and the industry yield data. When the income of the target resource management product changes with the income of the target industry, it means that the target resource management product is more sensitive to the target industry, and the position of the target industry in the target resource management product is relatively large.

[0171] In one implementation, correlation fitting is performed between the node yield data and the industry yield data corresponding to each measurement data collection time point, and the fitting parameters obtained by fitting are determined as the third sub-yield relationship data, including:

[0172] Obtain product daily yield sequence and industry yield sequence in multiple third nearest neighbor time periods corresponding to each measured data collection time point;

[0173] Perform linear fitting on the product daily yield sequence and the industry yield sequence to obtain the first linear fitting result;

[0174] Determine a first fitting parameter corresponding to each third nearest neighbor time period based on a slope coefficient of the first linear fitting result, and determine a second fitting parameter corresponding to each third nearest neighbor time period based on a residual square sum of the first linear fitting result;

[0175] The third sub-income relationship data is determined according to the first fitting parameter and the second fitting parameter.

[0176] The third nearest neighbor period may be a nearest neighbor period before the measurement data collection time point, for example, nearly one month, nearly two months, nearly three months, nearly six months, and nearly one year before the measurement data collection time point.

[0177] The product daily yield sequence can represent the daily yield of the target resource management product in the third nearest neighbor period, and the industry yield sequence can represent the daily yield of the target industry in the third nearest neighbor period. The product daily yield sequence can be obtained based on the first historical yield statistical data, and the daily yield of the target resource management product in the third nearest neighbor period can be calculated based on the first historical yield statistical data. The industry yield sequence can be obtained based on the second historical yield statistical data, and the daily yield of the target industry in the third nearest neighbor period can be calculated based on the second historical yield statistical data.

[0178] After obtaining the product daily yield sequence and the industry yield sequence in the third nearest neighbor period, a linear fit can be performed on the product daily yield sequence and the industry yield sequence. The first linear fitting result can represent the relative change relationship between the product daily yield and the industry yield in the third nearest neighbor period. Fig. 9The figure shows the first linear fitting result between the product yield and the industry yield in the third nearest neighbor period. The points in the figure represent the corresponding relationship between the daily product yield and the industry yield, and the straight line represents the first linear fitting result, which can reflect the overall change relationship between the product yield and the industry yield.

[0179] The first fitting parameter may be the slope coefficient of the first linear fitting result. When the industry yield is the independent variable and the product yield is the dependent variable, a higher slope coefficient of the first linear fitting result indicates that the product yield of the target resource management product is more sensitive to changes in the industry yield of the target industry. For example, the industry yield changes by only 1%, and the product yield changes by 3%. This can be explained that the target industry has a greater impact on the target resource management product, that is, the target industry occupies a higher position in the target resource management product. Conversely, a lower slope coefficient of the first linear fitting result indicates that changes in the industry yield of the target industry have less impact on the product yield of the target resource management product. For example, the industry yield changes by 1%, and the product yield changes by only 0.2%. At this point, the position occupied by the target industry in the target resource management product is so small that it is difficult to affect the overall yield change of the target resource management product.

[0180] The second fitting parameter may be the residual sum of squares of the first linear fitting result. The residual sum of squares may represent the error between the true value corresponding to the product yield and the industry yield and the predicted value corresponding to the first linear fitting result. The smaller the residual sum of squares, the closer the first linear fitting result is to the true value corresponding to the product yield and the industry yield, the higher the accuracy of the first linear fitting result in describing the relative change relationship between the product yield and the industry yield, and the higher the possibility that the target industry is included in the target resource management product.

[0181] The third sub-income relationship data may include first fitting parameters and second fitting parameters corresponding to the first linear fitting results in multiple third neighboring time periods corresponding to the measurement data collection time point.

[0182] After linear fitting is performed on the product yield corresponding to the target resource management product in the third nearest neighbor period corresponding to the measurement data collection time point and the industry yield corresponding to the target industry, the third sub-income relationship data is determined based on the slope coefficient corresponding to the linear fitting result and the residual square method. In this way, the third sub-income relationship data can describe the possibility of including the target industry in the target resource management product and the sensitivity of the target resource management product to the target industry from multiple third nearest neighbor periods corresponding to the measurement data collection time point, thereby improving the accuracy of determining the relative change relationship between the target resource management product and the target industry.

[0183] After determining the first sub-revenue relationship data, the second sub-revenue relationship data, and the third sub-revenue relationship, the revenue relationship data corresponding to each measurement data collection time point may be determined. That is, the revenue relationship data corresponding to each measurement data collection time point may be composed of the first sub-revenue relationship data, the second sub-revenue relationship data, and the third sub-revenue relationship data. The data dimension of the revenue relationship data may be equal to the sum of the dimensions of the first sub-revenue relationship data, the second sub-revenue relationship data, and the third sub-revenue relationship data.

[0184] Based on the first sub-revenue relationship data, the second sub-revenue relationship data and the third sub-revenue relationship data, the revenue relationship data can be determined from three aspects: the revenue difference between the target resource management product and the target industry, the correlation of revenue trends and the relative change relationship. It fully indicates the relationship between the target resource management product revenue and the target industry, and improves the accuracy of calculating the resource management product industry position information based on the revenue relationship data.

[0185] After determining the revenue relationship data corresponding to the multiple measurement data collection time points, the revenue relationship data can be sorted according to the time sequence of the multiple measurement data collection time points to obtain the revenue relationship data sequence corresponding to the target resource management product. Fig.10 As shown, the corresponding benefit relationship data D1, benefit relationship data D2, benefit relationship data D3 and benefit relationship data D4 are sorted in chronological order at the measured data collection time point T1, the measured data collection time point T2, the measured data collection time point T3 and the measured data collection time point T4 to obtain the corresponding benefit relationship data sequence.

[0186] By calculating the yield relationship data based on the node yield data and industry yield data corresponding to multiple measurement data collection time points, we can fully obtain the relationship between the yield data of the target resource management product and the yield data of the target industry at each measurement data collection time point, thereby improving the accuracy of determining the yield relationship data corresponding to the measurement data collection time point.

[0187] Step 540: Call the target neural network model to perform position calculation on the revenue relationship data sequence to obtain the position information corresponding to the target industry in the target resource management product corresponding to the calculation time information.

[0188] The target neural network model can be used to analyze the income relationship data corresponding to each measurement data collection time point in the income relationship data sequence, as well as the impact of the strong dependency information of multiple income relationship data in time sequence on the target industry position information.

[0189] In one embodiment, the training process of the target neural network model includes:

[0190] Obtain multiple training sample data;

[0191] The sample return relationship data sequence in multiple training sample data is used as the model input, and the corresponding industry position label sequence is used as the model training label training preset recurrent neural network model to obtain the trained target neural network model.

[0192] The training sample data includes a sample return relationship data sequence between sample resource management products and sample industries and a corresponding industry position label sequence.

[0193] The sample resource management product may be a resource management product for training a target neural network model. The sample industry may be one of the industries included in the sample resource management product. When the sample resource management product corresponds to one industry, then the industry may be used as the sample industry. When the sample resource management product corresponds to multiple industries, then each industry may be used as a sample industry.

[0194] The sample income relationship data sequence is a sequence of sample income relationship data corresponding to multiple sample time nodes in chronological order. The sample income relationship data is the data related to income of sample resource management products and sample industries at sample time nodes.

[0195] The industry position label sequence is a sequence of industry position labels corresponding to multiple sample time nodes in chronological order. The industry position label is the industry position information corresponding to the sample industry in the sample resource management product at the corresponding sample time sample node. For example, on XX / XX / XXXX, the industry position occupied by the sample industry in the sample resource management product is 30%, so the corresponding industry position label is 30%. The sample return relationship data sequence and the industry position label sequence correspond one-to-one in the order of multiple sample time nodes. For example, the sample time nodes include time T1, time T2, time T3 and time T4, and these four sample time nodes are arranged in chronological order, then the sample return relationship data sequence arranges the corresponding sample return relationship data in the above chronological order, and the industry position label sequence also arranges the corresponding industry position labels in the above chronological order.

[0196] Before obtaining multiple training sample data, the sample resource management products used for training can be determined first. The sample resource management products can be resource management products that can be selected by the object within a predetermined time period before the current time point. This is because the benefits of resource management products are greatly affected by the external environment. Therefore, using resource management products within a predetermined time period before the current time point to train the target neural network model allows the model to learn the environmental factors that affect the benefits of resource management products in the recent period of time, thereby improving the accuracy of industry position measurement of resource management products. In addition, the multiple sample resource management products corresponding to the multiple training sample data can cover multiple different industries, and for the same industry, the corresponding position information in different sample resource management products can be different. In this way, through multiple training sample data, the model can learn the connection between multiple industries and the corresponding income relationship data, as well as the connection between the position information of different industries corresponding to the same industry and the income relationship data, thereby improving the comprehensiveness of the target neural network model for position measurement of multiple industries, and the accuracy of industry position measurement.

[0197] In one implementation, a plurality of training sample data are obtained, the training sample data including a sample revenue relationship data sequence between a sample resource management product and a sample industry and a corresponding industry position label sequence, the sample revenue relationship data sequence being a sequence of sample revenue relationship data corresponding to a plurality of sample time nodes in chronological order, including:

[0198] Obtain the third historical income statistics and historical industry position information statistics of multiple sample resource management products within a preset period, and obtain the industry income statistics corresponding to resource management products in multiple industries within a preset period;

[0199] Determine multiple sample time nodes within a preset period, and determine sample industries among multiple industries;

[0200] Extracting a sample income relationship data sequence corresponding to each sample resource management product from the third historical income statistical data and the industry income statistical data based on multiple sample time nodes and sample industries;

[0201] Determine the industry position label sequence corresponding to each sample return relationship data sequence based on multiple sample time nodes and historical industry position information statistical data;

[0202] A plurality of training sample data are determined based on a plurality of sample resource management products and a sample industry's sample return relationship data sequence and a corresponding industry position label sequence.

[0203] The third historical revenue statistical data of the sample resource management product within the preset period may be the revenue-related data of the sample resource management product at different times within the preset period, which is the same as the first historical revenue statistical data corresponding to the target resource management product in the aforementioned implementation manner, and will not be repeated here.

[0204] The historical industry position information statistics may be the industry position information statistics of the sample resource management product at different times within a preset period. For example, at time T1 within the preset period, the industry position information statistics of the sample resource management product include: industry I1: 30%, industry I2: 25%, industry I3: 45%; at time T2 within the preset period, the industry position information statistics of the sample resource management product include: industry I1: 20%, industry I2: 30%, industry I3: 40%, industry I4: 10%.

[0205] The preset time period may be a specific time range. In one embodiment, the preset time period may be a predetermined time before the current time point, for example, one week, one month, or one year before the current time point. Since the income of the sample resource management product may be affected by the external environment, the use of income statistics data for a period that is too far away from the current time has little reference value for the industry position calculation of the resource management product at the current time. Therefore, the use of third historical income statistics data and historical industry position information statistics data within a predetermined time before the current time point can improve the accuracy of the industry position calculation corresponding to the resource management product.

[0206] In one embodiment, the third historical revenue statistics and historical industry position information statistics are statistically recorded in the form of a resource management product revenue report. Therefore, the third historical revenue statistics and historical industry position information statistics of multiple sample resource management products within a preset period of time can be obtained by obtaining multiple resource management product revenue reports of multiple sample resource management products within a preset period of time. The resource management product revenue report may include the industry position information distribution and revenue of the corresponding sample resource management product over a period of time, which has been described in detail in one embodiment of step 520 and will not be repeated here.

[0207] The industry revenue statistics corresponding to the resource management products of an industry may be the revenue of the resource management products in the industry within a preset period. Since the industry revenue can be reflected by the revenue of the resource management products in the industry, the industry revenue statistics corresponding to multiple resource management products in the industry may describe the overall revenue of the industry within the preset period.

[0208] In one embodiment, obtaining industry revenue statistics corresponding to resource management products of multiple industries within a preset period includes: obtaining industry revenue statistics of resource management products corresponding to each of the multiple industries within the preset period in the industry;

[0209] For each of the multiple industries, a resource management product including the industry can be determined from the multiple resource management products, and the industry revenue statistics corresponding to the industry can be obtained from the industry revenue statistics of different industries corresponding to the resource management products. The specific process is the same as the process of obtaining the industry revenue statistics of multiple resource management products corresponding to the target industry in an implementation of step 520, and will not be repeated here.

[0210] After obtaining the third historical profit statistics, historical industry position information statistics, and industry profit statistics corresponding to resource management products in multiple industries of multiple sample resource management products within the preset time period, it is also possible to determine multiple sample time nodes within the preset time period and determine sample industries among multiple industries.

[0211] The multiple sample time nodes may be multiple random time points within a preset period, or multiple time nodes determined according to a preset period. The method for determining the multiple sample time nodes may be consistent with the method for determining the multiple measurement data collection time points corresponding to the measurement time information.

[0212] In one embodiment, the sample industry can be determined from multiple industries by obtaining the industries included in the sample resource management product, and the industries included in the sample resource management product can be determined as the sample industry. For example, the multiple industries include: industry I1, industry I2, industry I3, industry I4 and industry I5, and the industries corresponding to the sample resource management product include: industry I2, industry I3 and industry I5. Therefore, the above three industries can be used as sample industries and combined with the sample resource management product to construct multiple training sample data.

[0213] In another embodiment, determining a sample industry from a plurality of industries includes:

[0214] Obtain at least one candidate industry corresponding to the sample resource management product and the effective duration of the sample resource management product's position in each candidate industry;

[0215] The sample industries are determined from the candidate industries based on the effective duration.

[0216] The candidate industry can be all industries that the sample resource management product holds positions in. The effective duration of the candidate industry joining the sample resource management product can indicate the duration from the time when the sample resource management product starts holding positions in the candidate industry to the time when the training sample data is constructed. For example, if the sample resource management product starts holding positions in the candidate industry one week before the training sample data is constructed, then the corresponding effective duration of the candidate industry is one week.

[0217] After determining the effective duration corresponding to the candidate industry, the sample industry can be determined from the candidate industry based on the effective duration. When the effective duration corresponding to the candidate industry is short, the sample return relationship data between the sample resource management product and the candidate industry is less. Therefore, the target neural network model may find it difficult to capture the relationship between the sample return relationship data and the industry position information of the candidate industry based on the sample return relationship data sequence, which may result in a waste of training resources.

[0218] Determining sample industries from candidate industries based on effective duration can be done by screening effective duration using a predetermined threshold. For example, if the predetermined threshold is one month, when the effective duration exceeds one month, the corresponding candidate industry can be used as a sample industry; conversely, when the effective duration is less than one month, the corresponding candidate industry cannot be used as a sample industry.

[0219] Determining the sample industry based on the effective duration of the sample resource management product's holding of the candidate industry can ensure that sufficient sample resource management product and sample industry sample return relationship data can be obtained, which is conducive to improving the comprehensiveness of the training sample data and making full use of training resources to obtain the accurate connection between the sample return relationship data and the industry position information corresponding to the sample industry.

[0220] After determining multiple sample time nodes and sample industries, the sample income relationship data sequence corresponding to each sample resource management product can be extracted from the third historical income statistical data and the industry income statistical data based on the multiple sample time nodes and sample industries.

[0221] To extract the income relationship data sequence corresponding to each sample resource management product, the sample income relationship data corresponding to the sample time node can be extracted from the third historical income statistical data corresponding to the sample resource management product and the industry income statistical data corresponding to the sample industry, and the sample income relationship data sequence can be constructed according to the order of the sample time nodes.

[0222] In one embodiment, extracting a sample revenue relationship data sequence corresponding to each sample resource management product from the third historical revenue statistical data and the industry revenue statistical data based on multiple sample time nodes and sample industries includes:

[0223] Extract multiple sample node yield data from the third historical yield statistical data based on multiple sample time nodes and sample industries;

[0224] Extracting multiple sample industry yield data corresponding to sample industries and multiple sample time nodes from industry yield statistical data;

[0225] The sample return relationship data corresponding to each sample time node is calculated based on the sample node return rate data and the corresponding sample industry return rate data, and the sample return relationship data is sorted according to the time sequence of multiple sample time nodes to obtain the sample return relationship data sequence corresponding to the sample resource management product.

[0226] The process of obtaining multiple sample node yield data and multiple sample industry yield data corresponding to multiple sample time nodes, calculating the sample yield relationship data of the corresponding sample time nodes based on the sample node yield data and the corresponding sample industry yield data, and sorting the multiple sample yield relationship data to obtain a sample yield relationship data sequence is the same as the process of obtaining multiple node yield data and multiple industry yield data corresponding to multiple measurement data collection time points, calculating the yield relationship data corresponding to the corresponding measurement data collection time points based on the node yield data and the corresponding industry yield data, and sorting the multiple yield relationship data to obtain a yield relationship data sequence in one implementation of step 530, and will not be repeated here.

[0227] By calculating the sample yield relationship data based on the sample node yield data corresponding to the sample time nodes and the sample industry yield data, we can fully obtain the relationship between the sample resource management product yield and the industry yield of the sample industry at each sample time node, thereby improving the accuracy of determining the sample yield relationship data corresponding to the sample time nodes.

[0228] After extracting multiple sample relationship data sequences, the industry position label sequence corresponding to each sample revenue relationship data sequence can be determined based on multiple sample time nodes and historical industry position information statistics. Since the historical industry position information statistics include industry position information statistics of sample resource management products at different times during a preset period, multiple industry position information of sample resource management products in sample industries at multiple sample time nodes can be determined from the historical industry position information statistics, and the multiple industry position information can be used as multiple industry position labels corresponding to multiple sample time nodes. Arranging multiple industry position labels in the time order of the corresponding multiple sample time nodes can form an industry position label sequence.

[0229] After obtaining the sample revenue relationship data sequence and the corresponding industry position label sequence, multiple training sample data can be determined based on the sample revenue relationship data sequence of multiple sample resource management products and sample industries and the corresponding industry position label sequence. Specifically, the sample revenue relationship data sequence corresponding to the sample resource management product and the sample industry and the industry position label sequence can be stored correspondingly to obtain the training sample data corresponding to the sample resource management product and the sample industry.

[0230] In another embodiment, before determining a plurality of training sample data according to a plurality of sample resource management products and sample industry sample revenue relationship data sequences and corresponding industry position label sequences, the method further includes:

[0231] Perform linear interpolation processing on the sample income relationship data sequence corresponding to each sample resource management product and the corresponding industry position label sequence to obtain the first sub-enhanced sample data;

[0232] Data splicing is performed on the sample income relationship data sequences corresponding to multiple sample resource management products and the corresponding industry position label sequences to obtain the second sub-enhanced sample data;

[0233] The enhanced sample data is determined according to the first sub-enhanced sample data and the second sub-enhanced sample data.

[0234] The industry position information statistics of resource management products are published at a predetermined time, for example, every six months or every year. Therefore, the amount of real industry position information data that can be used for model training is relatively scarce. Therefore, in this embodiment, the industry position information can be enhanced to achieve the expansion of training sample data.

[0235] First, linear interpolation processing can be performed on the sample return relationship data sequence corresponding to each sample resource management product and the corresponding industry position label sequence to obtain the first sub-enhanced sample data.

[0236] Linear interpolation processing can perform interpolation numerical estimation based on the linear relationship between the two adjacent data points on the left and right of the point to be interpolated in the one-dimensional data sequence. For example, the sample time nodes corresponding to the sample return relationship data sequence are extracted according to the monthly cycle. In order to enhance the sample, the sample return relationship data extracted with a monthly cycle can be expanded to the sample return relationship data with a weekly cycle through linear difference processing. The sample return relationship data with a weekly cycle can be filled and interpolated based on the linear relationship corresponding to the monthly sample return relationship data.

[0237] In one implementation, a linear interpolation process is performed on the sample revenue relationship data sequence and the corresponding industry position label sequence corresponding to each sample resource management product to obtain first sub-enhanced sample data, including:

[0238] Obtain the first sampling time and first industry position information corresponding to each sample income relationship data in the sample income relationship data sequence corresponding to each sample resource management product;

[0239] Performing a linear fit on the first sampling time and the first industry position information to obtain a second linear fit result;

[0240] Perform interpolation processing based on the second linear fitting result to obtain multiple second sampling times and corresponding second industry position information;

[0241] The enhanced profit relationship data corresponding to each second sampling time is obtained, and the first sub-enhanced sample data is determined according to the enhanced profit relationship data and the second industry position information.

[0242] The first sampling time corresponding to the sample income relationship data may be the time when the sample income relationship data corresponding to the sample resource management product is collected, and may be the sample time node corresponding to the sample income relationship data in the aforementioned step. The first industry position information may be the industry position of the sample resource management product occupied by the sample industry at the first sampling time. Therefore, the first industry position information may be consistent with the industry position label corresponding to the sample time node in the industry position label sequence.

[0243] By performing a linear fit on the first sampling time and the first industry position information, a second linear fit result can be obtained. The second linear fit result can describe the change in the position of the sample industry in the sample resource management product over time. For example Fig.11 As shown, based on the linear relationship between the first sampling time T1-first sampling time T9 and the corresponding first industry position information, the second linear fitting result is obtained. It should be noted that the second linear fitting result can be the result of fitting two adjacent first sampling times, that is, the second linear fitting result can be a line graph composed of multiple straight lines.

[0244] After obtaining the second linear fitting result, multiple second sampling times can be obtained based on the first sampling time. The second sampling time can be the time between two adjacent first sampling times and in the second linear fitting result. Fig.11 In the method, a plurality of second sampling times are randomly determined between every two adjacent first sampling times. After the second sampling time is determined, the second industry position information corresponding to the second sampling time can be determined based on the second linear fitting result.

[0245] For example, the interpolation processing results are shown in Table 1.

[0246]

[0247] Table 1

[0248] In Table 1, the industry position information is published monthly, that is, the industry position information corresponding to the 1st week, the 4th week, and the 7th week is known. Enhancing the sample data can expand the monthly published industry position information to weekly industry position information. Therefore, based on the linear fitting results between the 1st week and the 4th week and the corresponding industry position information, it can be determined that the industry position information of the 2nd week is 3, and the industry position information of the 3rd week is 5. Based on the linear fitting results between the 4th week and the 7th week and the corresponding industry position information, it can be determined that the industry position information of the 5th week is 6.5, and the industry position information of the 6th week is 6.

[0249] After performing interpolation processing to obtain multiple second sampling times and corresponding second industry position information, the enhanced revenue relationship data corresponding to each second sampling time can be obtained. The enhanced revenue relationship data corresponding to the second sampling time can be obtained based on the third historical revenue statistics data and industry revenue statistics data corresponding to the sample resource management product. The specific process is similar to the process of obtaining the sample revenue relationship data corresponding to the sample time node in the aforementioned implementation method, and will not be repeated here.

[0250] The first sub-enhanced sample data is determined based on the enhanced profit relationship data and the second industry position information. The enhanced profit relationship data corresponding to the second sampling time and the second industry position information can be stored correspondingly to form the first sub-enhanced sample data corresponding to the second sampling time.

[0251] After linear fitting of the first sampling time and the first industry position information, the second linear fitting result is interpolated to obtain multiple second sampling times and corresponding second industry position information, thereby ensuring the accuracy of the linear interpolation processing, so that the first sub-enhanced sample data can conform to the true linear relationship between the sample resource management product and the industry position information corresponding to the sample industry, thereby improving the accuracy of sample data enhancement.

[0252] The second sub-enhanced sample data is obtained by splicing the sample return relationship data sequences corresponding to multiple sample resource management products and the corresponding industry position label sequences.

[0253] In one implementation, data splicing is performed on sample revenue relationship data sequences and corresponding industry position label sequences corresponding to multiple sample resource management products to obtain second sub-enhanced sample data, including:

[0254] Perform feature concatenation on the sample revenue relationship data sequences corresponding to multiple sample resource management products to obtain enhanced sample features;

[0255] The industry position label sequences corresponding to multiple sample resource management products are sequentially spliced ​​to obtain enhanced sample labels, and the second sub-enhanced sample data is determined according to the enhanced sample features and the enhanced sample labels.

[0256] The sample income relationship data sequences corresponding to multiple sample resource management products are subjected to feature splicing. By splicing the sample income relationship data sequences corresponding to multiple sample resource management products together head to tail, enhanced sample features can be obtained.

[0257] In one implementation, feature concatenation is performed on sample revenue relationship data sequences corresponding to multiple sample resource management products to obtain enhanced sample features, including:

[0258] Adding a connection variable to the end of each sample benefit relationship data in the sample benefit relationship data sequence corresponding to the plurality of sample resource management products;

[0259] Based on the connection variables, the sample income relationship data sequences corresponding to multiple sample resource management products are concatenated head to tail to obtain enhanced sample features.

[0260] The connection variable can be used to indicate whether the connection is the connection of the sample revenue relationship data sequence corresponding to the sample resource management product. That is to say, when the connection variable indicates that the connection is not the connection of the sample revenue relationship data sequence corresponding to the sample resource management product, the next digit connected by the connection variable is the next sample revenue relationship data in the same sample revenue relationship data sequence; and when the connection variable indicates that the connection is the connection of the sample revenue relationship data sequence corresponding to the sample resource management product, the next digit connected by the connection variable is the sample revenue relationship data sequence corresponding to another sample resource management product. The purpose of adding the connection variable is to distinguish the sample revenue relationship data sequences corresponding to different sample resource management products in the enhanced sample features. The purpose of adding the connection variable to the end of each sample revenue relationship data in a complete sample revenue relationship data sequence is to unify the dimensions of the sample revenue relationship data.

[0261] After adding a connection variable to the sample-income relationship data sequence, the sample-income relationship data sequences corresponding to multiple sample resource management products can be concatenated end to end to obtain enhanced sample features.

[0262] For example Fig.12As shown, when the connection variable indicates that the connection is the connection of the sample income relationship data in the sample income relationship data sequence, the connection variable is 0; when the connection variable indicates that the connection is the connection of the sample income relationship data sequence corresponding to the sample resource management product, the connection variable is 1. Among the multiple sample income relationship data in the sample income relationship data sequence S1 corresponding to the sample resource management product A1, the last connection of the sample income relationship data D1 and the sample income relationship data D2 is the connection of the sample income relationship data in the sample income relationship data sequence, so the connection variable 0 is added; the last connection of the income relationship data D3 is the connection of the sample income relationship data sequence S1 and the sample income relationship data sequence S2 corresponding to the sample resource management product A2, so the connection variable 1 is added. The same is true for adding a connection variable to the last position of the sample income relationship data in the sample income relationship data sequence S2.

[0263] By adding connection variables, the sample revenue relationship data sequences corresponding to different sample resource management products can be clearly distinguished in the enhanced sample features, so that the preset recurrent neural network model can quickly adapt to different sample revenue relationship data sequences and improve training accuracy. Fig.13 The figure shows the sensitivity of features at different positions in the enhanced sample features when the preset recurrent neural network model is learning based on the enhanced sample features obtained by data concatenation after adding connection variables. Fig.13 In the figure, the connection variables at position B1 and position B2 indicate that the sample income relationship data sequence has changed. It can be seen that the preset recurrent neural network model can sensitively distinguish the sample income relationship data sequences corresponding to different sample resource management products, and learn for different sample income relationship data sequences.

[0264] The industry position label sequences corresponding to multiple sample resource management products can also be spliced ​​by connecting the industry position label sequences corresponding to multiple sample resource management products end to end to obtain enhanced sample labels. It should be noted that the splicing order of the industry position label sequences corresponding to multiple sample resource management products is consistent with the splicing order of the corresponding sample return relationship data sequences.

[0265] In one implementation, the industry position label sequences corresponding to multiple sample resource management products are concatenated to obtain enhanced sample labels, including:

[0266] Add a connection variable to the end of each industry position label in the industry position label sequence corresponding to multiple sample resource management products;

[0267] Based on the connection variables, the industry position label sequences corresponding to multiple sample resource management products are concatenated head to tail to obtain enhanced sample labels.

[0268] Similar to the implementation in the aforementioned steps, the connection variable can be used to indicate whether the connection is the connection of the industry position label sequence corresponding to the sample resource management product. That is, when the connection variable indicates that the connection is not the connection of the industry position label sequence corresponding to the sample resource management product, the next bit connected by the connection variable is the next industry position label in the same industry position label sequence; and when the connection variable indicates that the connection is the connection of the industry position label sequence corresponding to the sample resource management product, the next bit connected by the connection variable is the industry position label sequence corresponding to another sample resource management product.

[0269] After adding connection variables to the industry position label sequence, the industry position label sequences corresponding to multiple sample resource management products can be spliced ​​head to tail to obtain enhanced sample labels. Adding connection variables to the industry position label sequence and splicing to obtain enhanced sample labels allows the preset recurrent neural network model to distinguish the industry position label sequences corresponding to different sample resource management products, thereby improving the accuracy of model training.

[0270] After the enhanced sample features and the enhanced sample labels are obtained, the enhanced sample features and the enhanced sample labels may be stored correspondingly to form second sub-enhanced sample data.

[0271] By performing feature splicing on the sample return relationship data sequences corresponding to multiple sample resource management products and performing sequence labeling on the industry position label sequences, it is possible to expand the data volume of the training sample data. By increasing the data volume of the training sample data, the robustness of the trained target neural network model can be improved.

[0272] After the first sub-enhanced sample data is obtained by linear interpolation processing and the second sub-enhanced sample data is obtained by data splicing, the first sub-enhanced sample data and the second sub-enhanced sample data can be used together as enhanced sample data.

[0273] Based on the above implementation, multiple training sample data are determined according to multiple sample resource management products and sample industry sample revenue relationship data sequences and corresponding industry position label sequences, including:

[0274] Multiple training sample data are determined based on enhanced sample data, sample return relationship data sequences of multiple sample resource management products and sample industries, and corresponding industry position label sequences.

[0275] Specifically, the sample revenue relationship data sequence corresponding to the sample resource management product and the sample industry can be stored in correspondence with the industry position label sequence to obtain the known training sample data corresponding to the sample resource management product and the sample industry. Multiple known training data corresponding to multiple sample resource management products and sample industries and enhanced sample data can together constitute multiple training sample data.

[0276] Enhanced sample data is obtained through linear interpolation processing and data splicing. The enhanced sample data can be used to expand the training sample data, which solves the problem of insufficient training sample data and improves the comprehensiveness of the use of training sample data, thereby improving the accuracy of target neural network model training in subsequent steps.

[0277] In the above implementation, based on the third historical revenue statistics of the sample resource management product in the preset period and the industry revenue statistics corresponding to the sample industry, the sample revenue relationship data corresponding to the multiple sample time nodes are determined, thereby forming a sample revenue relationship data sequence, and the industry position label corresponding to each sample revenue relationship data is determined from the historical industry position information statistics, thereby forming an industry position label sequence. In this way, accurate sample revenue relationship data and industry position label data corresponding to the sample resource management product and the sample industry can be obtained, which improves the authenticity and reliability of the obtained training sample data.

[0278] The preset recurrent neural network model can be used to process time series data and can capture the strong dependencies between time series data.

[0279] In one embodiment, the preset recurrent neural network model may be a long short-term memory network (LSTM). LSTM can handle the problems of gradient vanishing and gradient exploding, and while handling time series data well, it can overcome the problem that general recurrent neural network models are difficult to train.

[0280] LSTM includes forget gate, input gate and output gate. The forget gate determines the information from the previous neuron that the current neuron needs to retain or discard; the input gate determines the information added to the current neuron; the output gate determines which of the updated information of the current neuron is output to the next neuron. The specific structure is as follows Fig.14 As shown, h t-1 is the output vector of the previous neuron, x t is the input vector of the current neuron, C t-1 is the neuron state vector of the previous neuron. The forget gate can be mapped through a sigmoid nonlinear mapping based on h t-1 With x t Output the first intermediate vector. The value of each dimension in the first intermediate vector is between 0 and 1, 1 means completely retained, and 0 means completely discarded.t-1 Multiplication can determine the information retained and discarded in the neuron state of the previous neuron. The input gate is activated by a sigmoid function based on h t-1 With x t Output the second intermediate vector, which can be used to determine the value that needs to be updated in the current neuron. The first neural layer can be passed through a tanh layer based on h t-1 With x t Create a candidate value vector for the current neuron. Multiply the second intermediate vector by the candidate value vector for the current neuron to determine the neuron state vector for the current neuron. Add the neuron state vector for the current neuron to the information retained and discarded in the neuron state of the previous neuron to obtain the final neuron state vector C for the current neuron. t The output layer can be mapped through a sigmoid nonlinear mapping based on h t-1 With x t Determine what information can be output, and use the second neural layer to calculate the final neuron state vector C t Process it and multiply the processing result with the output result of the output layer to get the output vector h of the current neuron t .

[0281] Through the above data processing process of neurons in LSTM, it can be known that in the embodiment of the present disclosure, the sample return relationship data sequence is input into the LSTM model, and each neuron in the LSTM model inputs a sample return relationship data in the sample return relationship data sequence, and outputs the industry position feature data corresponding to the sample return relationship data. In addition to the first neuron, each subsequent neuron will also receive the output result and neuron memory information from the previous neuron to obtain the dependency relationship between the sample return relationship data.

[0282] like Fig.15 As shown, the sample return relationship data D1-sample return relationship data D4 in the sample return relationship data sequence are used as the input vectors of neurons N1-neuron N4 in LSTM in sequence. Neuron N1 obtains the corresponding output industry position information K1 based on the sample return relationship data D1, and inputs the output result and the neuron state vector into neuron N2. Neuron N2 obtains the corresponding output industry position information K2 based on the parameters transmitted by neuron N1 and the sample return relationship data D2, and inputs the output result and the neuron state vector into neuron N3. Neuron N3 obtains the corresponding output industry position information K3 based on the parameters transmitted by neuron N2 and the sample return relationship data D3, and inputs the output result and the neuron state vector into neuron N4. Neuron N4 obtains the corresponding output industry position information K4 based on the parameters transmitted by neuron N3 and the sample return relationship data D4.

[0283] In another embodiment, the preset recurrent neural network model includes an LSTM layer and a fully connected layer. The LSTM layer can be responsible for determining the industry position feature data corresponding to each sample return relationship data and mapped to the feature space based on the sample return relationship data sequence, and the fully connected layer can map multiple industry position feature data corresponding to multiple sample return relationship data to the tag space to obtain multiple industry position information output values. The structure of the preset recurrent neural network is as follows: Fig.16 As shown, the sample return relationship data sequence is input into the LSTM layer, and then the output vector is input into the fully connected layer to obtain an industry position output sequence composed of multiple industry position information output values.

[0284] The industry position label sequence corresponding to the sample return relationship data sequence is used as the model training label to guide the parameter optimization direction of the preset recurrent neural network model. Therefore, the training loss between the industry position label sequence and the industry position output sequence can be calculated. The training loss can be calculated using the mean absolute error (MAE). The preset recurrent neural network model is trained based on the training loss to obtain the target neural network model.

[0285] In one embodiment, a preset recurrent neural network model is trained using a sample return relationship sequence in a plurality of training sample data as a model input, and a corresponding industry position label sequence is used as a model training label training to obtain a trained target neural network model, including:

[0286] Dividing the plurality of training sample data into a first training set and a first validation set;

[0287] Randomly extract a preset proportion of target training sample data from the first training set, and construct a second validation set with the extracted target training sample data and the first validation set, and construct a second training set with the training sample data other than the target training sample data in the first training set;

[0288] A preset long short-term memory network model is trained based on the sample return relationship data sequence in the second training set as the model input and the corresponding industry position label sequence as the model training label;

[0289] The trained long short-term memory network model is verified according to the second verification set, and when the verification result meets the preset verification condition, the trained long short-term memory network model is determined as the target neural network model.

[0290] The plurality of training sample data are divided into a first training set and a first validation set, for example, at a ratio of 8:2.

[0291] Extract target training sample data from the first training set according to a preset ratio, and use the target training sample data and the first validation set to form a second validation set; use other samples other than the target training sample data to form a second training set. For example, extract 1 / 5 of the samples from the first training set as the target training sample data, and use them to form a second validation set, and use the remaining 4 / 5 of the samples in the first training set to form a second training set.

[0292] The sample return relationship data sequence in the second training set is used as the model input, and the corresponding industry position label sequence is used as the model training label to train the preset long short-term memory network model. The training process of the long short-term memory network model has been described in detail in the above implementation method and will not be repeated here.

[0293] After the long short-term memory network model is trained, the model can be verified using the second verification set. When the verification result meets the preset verification condition, the long short-term memory network model can be determined as the target neural network model. The preset verification condition can be that the prediction accuracy reaches a predetermined threshold, etc.

[0294] In one embodiment, a preset proportion of target training sample data is randomly extracted from a first training set, and a second validation set is constructed by combining the extracted target training sample data with the first validation set, and a second training set is constructed by combining the training sample data other than the target training sample data in the first training set, including: determining a predetermined number of training times, and for each training, randomly extracting a preset proportion of target training sample data from the first training set, and a second validation set is constructed by combining the extracted target training sample data with the first validation set, and a second training set is constructed by combining the training sample data other than the target training sample data in the first training set;

[0295] Based on the sample return relationship data sequence in the second training set as the model input, and the corresponding industry position label sequence as the model training label training of the preset long short-term memory network model, including: for each training, based on the corresponding sample return relationship data sequence in the second training set as the model input, and the corresponding industry position label sequence as the model training label training of the preset long short-term memory network model;

[0296] The trained long short-term memory network model is verified according to the second verification set. When the verification result meets the preset verification condition, the trained long short-term memory network model is determined as the target neural network model, including:

[0297] For each training, the trained LSTM network model is verified based on the corresponding second validation set;

[0298] When the verification results of the predetermined number of training times meet the preset verification conditions, the long short-term memory network model obtained through training is determined as the target neural network model.

[0299] That is, the first training set is split multiple times according to the predetermined number of training times, and for each training, the corresponding second training set is used for training, and the corresponding second verification set is used for verification. When the multiple verification results can meet the preset verification conditions, the long short-term memory network model obtained by training can be determined as the target neural network model.

[0300] like Fig.17 As shown, the training sample data set is divided into a first training set and a first validation set. The predetermined number of training times is 5, and the target training sample data is extracted from the first training set 5 times. For each training, the target training sample data and the first validation set constitute a second validation set, and the data other than the target training sample data in the first training set is constructed as a second training set. The long short-term memory network model is trained using the second training set, and then the long short-term memory network model is validated using the second validation set. When the validation results after 5 trainings can meet the preset validation conditions, the trained long short-term memory network model is determined as the target neural network model.

[0301] Dividing the second training set and the second validation set based on the first training set can improve the generalization ability of the preset recurrent neural network model, so that the model can be trained and validated on different data subsets, thereby improving the accuracy of evaluating unknown data and reducing the risk of model overfitting.

[0302] In one embodiment, Fig.18 As shown in the figure, the training process of the target neural network model includes:

[0303] Step 1810: Obtain third historical revenue statistics and historical industry position information statistics for multiple sample resource management products within a preset time period, and obtain industry revenue statistics corresponding to resource management products in multiple industries within the preset time period.

[0304] The third historical income statistics of the sample resource management product in the preset period may be income-related data of the sample resource management product at different times in the preset period. The historical industry position information statistics may be industry position information statistics of the sample resource management product at different times in the preset period.

[0305] The third historical revenue statistics data and historical industry position information statistics data of multiple sample resource management products within a preset time period can be obtained by obtaining multiple resource management product revenue reports of multiple sample resource management products within the preset time period. The resource management product revenue report may include the industry position information allocation and revenue of the corresponding sample resource management products over a period of time. The resource management product revenue report may include the revenue of the sample resource management products over a period of time and the positions occupied by multiple industries in the sample resource management products. Depending on the preset time period, the resource management product revenue report corresponding to the target resource management product and whose statistical time includes the preset time period can be obtained, and then the third historical revenue statistics data and historical industry position information statistics data can be extracted from the resource management product revenue report.

[0306] The industry revenue statistics may be the revenue situation of the resource management products that include the industry in a preset time period. For each of the multiple industries, the resource management products that include the industry can be determined from the multiple resource management products, and the industry revenue statistics corresponding to the industry can be obtained from the industry revenue statistics of different industries corresponding to the resource management products. The resource management product revenue report may not only count the revenue data of the corresponding resource management product, but may also include industry position information data of multiple industries and independent industry revenue statistics of each industry. Therefore, the industry revenue statistics corresponding to the industry may be extracted from the resource management product revenue reports corresponding to the multiple resource management products that include the industry.

[0307] Step 1820: extract the sample profit relationship data sequence corresponding to each sample resource management product from the third historical profit statistical data and the industry profit statistical data based on multiple sample time nodes and sample industries, and determine the industry position label sequence corresponding to each sample profit relationship data sequence based on multiple sample time nodes and historical industry position information statistical data.

[0308] To extract the income relationship data sequence corresponding to each sample resource management product, the sample income relationship data corresponding to the sample time node can be extracted from the third historical income statistical data corresponding to the sample resource management product and the industry income statistical data corresponding to the sample industry, and the sample income relationship data sequence can be constructed according to the order of the sample time nodes.

[0309] The sample return relationship data can be determined based on the sample node return rate data of the sample resource management products at multiple sample time nodes and the sample industry return rate data of the sample industry at multiple sample time nodes.

[0310] The sample node yield data may represent the yield data of the sample resource management product at the corresponding sample time node, and may be extracted from the third historical yield statistical data corresponding to the sample resource management product based on the sample time node. The sample industry yield data may represent the yield data of the sample industry at the corresponding sample time node, and may be extracted from the industry yield statistical data corresponding to the sample industry based on the sample time node.

[0311] After extracting the sample node yield data and the sample industry yield data, the sample yield relationship data corresponding to each sample time node can be calculated based on the sample node yield data and the corresponding sample industry yield data. The sample yield relationship data may include the difference between the sample node yield data corresponding to the corresponding sample time node and the sample industry yield data, the correlation coefficient between the sample yield trend corresponding to the sample node yield data corresponding to the sample time node and the industry yield trend corresponding to the sample industry yield data, and the fitting parameters after the correlation fitting of the sample node yield data and the sample industry yield data.

[0312] After obtaining the sample-income relationship data corresponding to the multiple sample time nodes, the multiple sample-income relationship data may be sorted according to the time sequence of the multiple sample time nodes to form a sample-income relationship data sequence.

[0313] After extracting multiple sample relationship data sequences, the industry position label sequence corresponding to each sample revenue relationship data sequence can be determined based on multiple sample time nodes and historical industry position information statistical data. Specifically, multiple industry position information of sample resource management products in sample industries at multiple sample time nodes can be determined from historical industry position information statistical data, and the multiple industry position information can be used as multiple industry position labels corresponding to multiple sample time nodes. Arranging multiple industry position labels in the time order of the corresponding multiple sample time nodes can form an industry position label sequence.

[0314] Step 1830: Perform linear interpolation processing on the sample return relationship data sequence corresponding to each sample resource management product and the corresponding industry position label sequence to obtain the first sub-enhanced sample data.

[0315] Linear interpolation processing can perform interpolation numerical estimation based on the linear relationship between the two adjacent data points on the left and right of the point to be interpolated in the one-dimensional data sequence. For example, the sample time nodes corresponding to the sample income relationship data sequence are extracted according to the monthly cycle. In order to enhance the sample, the sample income relationship data extracted with a monthly cycle can be expanded to the sample income relationship data with a weekly cycle through linear difference processing. The sample income relationship data with a weekly cycle can be filled and interpolated based on the linear relationship corresponding to the monthly income relationship data.

[0316] Therefore, the linear interpolation processing of the sample return relationship data sequence can be performed by linearly fitting the industry position labels corresponding to the two adjacent sample time nodes, and determining the intermediate sampling time node from the two adjacent sample time nodes, and determining the enhanced industry position label corresponding to the intermediate sampling time node based on the linear fitting result. Based on the third historical return statistical data and industry return statistical data corresponding to the sample resource management product, the enhanced return relationship data corresponding to the intermediate sampling time node is obtained. The enhanced return relationship data and the enhanced industry position label are stored in correspondence, which can constitute the first sub-enhanced sample data corresponding to the intermediate sampling time node.

[0317] Step 1840: perform data splicing on the sample revenue relationship data sequences corresponding to the multiple sample resource management products and the corresponding industry position label sequences to obtain the second sub-enhanced sample data.

[0318] First, feature splicing of multiple sample return relationship data sequences can be performed to obtain enhanced sample features; then, data splicing of multiple industry position label sequences can be performed in the order of splicing of the corresponding multiple sample return relationship data sequences to obtain enhanced sample labels; the enhanced sample features and enhanced sample labels can constitute the second sub-enhanced sample data.

[0319] Feature concatenation of multiple sample-revenue relationship data sequences can be performed by adding a connection variable to the end of each sample-revenue relationship data in each sample-revenue relationship data sequence, and concatenating the multiple sample-revenue relationship data sequences head to tail based on the connection variable to obtain enhanced sample features.

[0320] The connection variable can be used to indicate whether the connection is the connection of the sample income relationship data sequence corresponding to the sample resource management product. When the connection variable indicates that the connection is not the connection of the sample income relationship data sequence corresponding to the sample resource management product, the next bit connected by the connection variable is the next sample income relationship data in the same sample income relationship data sequence; when the connection variable indicates that the connection is the connection of the sample income relationship data sequence corresponding to the sample resource management product, the next bit connected by the connection variable is the sample income relationship data sequence corresponding to another sample resource management product.

[0321] Multiple industry position label sequences can also be concatenated by adding a connection variable to the end of each industry position label in the industry position label sequences corresponding to multiple sample resource management products, and concatenating the industry position label sequences corresponding to the multiple sample resource management products based on the connection variable to obtain enhanced sample labels.

[0322] Step 1850: determine the target training sample data based on the first sub-enhanced sample data, the second sub-enhanced sample data, a plurality of sample resource management products and sample industry sample revenue relationship data sequences, and a corresponding industry position label sequence.

[0323] The sample revenue relationship data sequences of multiple sample resource management products and sample industries and the corresponding industry position label sequences are stored correspondingly to form known training sample data. The first sub-enhanced sample data and the second sub-enhanced sample data can be used as enhanced sample data. The known training sample data and the enhanced sample data can be used together as target training sample data.

[0324] Step 1860: Train a preset long short-term memory network model based on the target training sample data to obtain a trained target neural network model.

[0325] The sample return relationship data sequence in the target training sample data is used as the model input, and the industry position label sequence is used as the model training label to train the preset long short-term memory network model.

[0326] Each neuron in the long short-term memory network model can input a sample return relationship data in the sample return relationship data sequence. Except for the first neuron, each subsequent neuron will also receive the output result and neuron memory information from the previous neuron to obtain the dependency relationship between the sample return relationship data. Based on the sample return relationship data in the output neuron and the dependency relationship between the sample return relationship data, the industry position feature data corresponding to each sample return relationship data and mapped to the feature space is determined.

[0327] The industry position feature data corresponding to each sample return relationship data is mapped to the label space using a fully connected layer to obtain the industry position information output value. Multiple industry position information output values ​​corresponding to multiple sample return relationship data can constitute an industry position information output value sequence. The model training loss is calculated based on the industry position information output value sequence and the industry position label sequence. Specifically, the model training loss can be determined by calculating the mean absolute error between the industry position information output value sequence and the industry position label sequence. The preset long short-term memory network model is trained based on the model training loss.

[0328] When training the preset long short-term memory network model, the target training sample data can be divided into a first training set and a first validation set according to a predetermined ratio, and the first training set and the first validation set can be used to perform a predetermined number of training times.

[0329] For each training, a predetermined ratio of training sample data in the first training set can be used as the target training sample data, and the training sample data and the first validation set can constitute the second validation set. The training sample data in the first training set except the target training sample data constitute the second training set. For example, the long short-term memory network model is trained 5 times, and the first training set is evenly divided into 5 parts. Before each training, one of the parts and the first validation set constitute the second validation set, and the remaining 4 parts constitute the second training set. In each training, the preset long short-term memory network model is trained by the above process using the training sample data in the corresponding second training set.

[0330] After each training, the trained long short-term memory network model can be verified using the second verification set corresponding to this training, and the verification result can be recorded. When the verification result corresponding to the predetermined number of training times meets the preset verification condition, the trained long short-term memory network model can be determined as the target neural network model. For example, the predetermined verification condition is that the average value of the verification results corresponding to the predetermined number of training times reaches a predetermined threshold.

[0331] The preset recurrent neural network model is trained using the sample revenue relationship data of the sample resource management products and sample industries at multiple sample time nodes and the industry position labels to obtain the target neural network model. Since the recurrent neural network model can be used to deal with data dependency problems in time-related sequences, through the above training process, the target neural network model can learn the relationship between the sample revenue relationship data and the industry position labels corresponding to multiple sample time nodes, so that the target neural network model has the ability to predict industry position information based on the revenue relationship data sequence. Through the above training method, the accuracy of the target neural network model in position measurement is improved.

[0332] The target neural network model is called to perform position measurement on the revenue relationship data sequence corresponding to the measurement time information, and the position information corresponding to the target industry in the target resource management product corresponding to the measurement time information is obtained.

[0333] The process of the target neural network model performing position calculation based on the return relationship data sequence is the same as the process of the preset recurrent neural network model performing position calculation based on the sample return relationship data sequence as input in the aforementioned implementation, and will not be repeated here.

[0334] In summary, the disclosed embodiment determines the revenue relationship data sequence based on multiple measurement data collection time points corresponding to the measurement time information, the first historical revenue statistical data corresponding to the target resource management product, and the second historical revenue statistical data corresponding to the target industry. Through the revenue relationship data sequence, the target neural network model can analyze the first historical revenue statistical data and the second historical revenue statistical data corresponding to multiple measurement data collection time points to predict the position information corresponding to the target industry in the target resource management product at the measurement time. Therefore, in the position calculation method of the resource management product of the disclosed embodiment, the target neural network model can use the revenue-related information of the target resource management product and the target industry collected at multiple measurement data collection time points in the revenue relationship data sequence, as well as the long-term dependence of the revenue-related information in time series as a basis to predict the position information corresponding to the target industry at the measurement time, thereby improving the accuracy of calculating the industry position corresponding to the resource management product.

[0335] The embodiments of the present disclosure are described in detail in combination with specific application scenarios.

[0336] like Fig.19 FIG. 1 is another flow chart of the method for calculating the position of the resource management product provided by the present disclosure. The method specifically includes the following steps:

[0337] Step 1911, determine the sample resource management product, the sample industry and multiple sample time nodes, and determine the industry position label sequence corresponding to the sample industry at the multiple sample time nodes.

[0338] The sample resource management product may be a resource management product available for selection by the object within a predetermined time period before the time point at which the resource management product is measured. The sample industry may be any industry included in the sample resource management product. The sample time node may be a plurality of predetermined times within the predetermined time period, for example, the predetermined time period is one month before the current time point, and the sample time point is a time node corresponding to each Monday in a month.

[0339] The industry position label indicates the industry position information occupied by the sample industry in the sample resource management product at the corresponding sample time sample node. The industry position label sequence is a sequence of industry position labels corresponding to multiple sample time nodes in chronological order.

[0340] Industry position labels can be obtained based on historical industry position information statistics of sample resource management products. Historical industry position information statistics can record sample income-related data and industry position information statistics corresponding to multiple time nodes of sample resource management products within a predetermined period. Therefore, multiple industry position information of sample resource management products in sample industries at multiple sample time nodes can be determined from historical industry position information statistics, and multiple industry position information can be used as multiple industry position labels corresponding to multiple sample time nodes. Arranging multiple industry position labels in the time order of the corresponding multiple sample time nodes can form an industry position label sequence.

[0341] Step 1912: Obtain product yield values ​​and industry yield values ​​corresponding to multiple first neighboring time periods corresponding to each sample time node, and calculate the yield difference corresponding to the corresponding first neighboring time period based on the product yield value and the industry yield value.

[0342] The plurality of first neighboring time periods corresponding to the sample time node may be a preset neighboring time period before the sample time node, for example, one day or one week before the sample time node.

[0343] The yield difference corresponding to each first neighboring time period may be the difference between the product yield value and the industry yield value corresponding to the first neighboring time period.

[0344] The product yield value within the first nearest neighbor period can be obtained based on the historical yield statistical data corresponding to the sample resource management products. The historical yield statistical data can record the yield data corresponding to the sample resource management products at multiple time points within the preset period. Based on the starting net income value and the ending net income value in the first nearest neighbor period in the historical yield statistical data, the product yield value of the first nearest neighbor period can be calculated. The industry yield value can be obtained based on the industry yield statistical data of the sample industry. The method of determining the industry yield value corresponding to the first nearest neighbor time period based on the industry yield statistical data is similar to the method of calculating the product yield value based on the historical yield statistical data, which will not be repeated here.

[0345] The difference between the product yield value and the industry yield value corresponding to each first neighbor period is calculated to obtain the yield difference. The yield difference can reflect the difference between the sample resource management product and the sample industry. The smaller the difference, the larger the industry position of the sample industry in the sample resource management product.

[0346] Step 1913, obtain the product daily return trend data and the industry daily return trend data in multiple second nearest neighbor time periods corresponding to each sample time node, and calculate the return trend correlation coefficient of the corresponding second nearest neighbor time period based on the product daily return trend data and the industry daily return trend data.

[0347] The second neighboring time period may be a neighboring time period before the sample time node, for example, nearly one month, nearly two months, nearly three months, nearly six months, nearly one year, etc. before the sample time node.

[0348] The daily product income trend data within the second nearest neighbor period can be obtained based on the historical income statistical data corresponding to the sample resource management product. The historical income statistical data can record the income data of the sample resource management product at multiple time points within the preset period. The daily product income trend data within the second nearest neighbor period is generated by obtaining the daily income data changes of the sample resource management product within the second nearest neighbor period from the historical income statistical data. The industry daily income trend data can be obtained based on the industry income statistical data of the sample industry. The method of determining the industry daily income trend data corresponding to the second nearest neighbor time period based on the industry income statistical data is similar to the method of obtaining the product daily income trend data based on the historical income statistical data, which will not be repeated here.

[0349] Calculate the Pearson correlation coefficient between the product daily return trend data and the industry daily return trend data corresponding to each second nearest neighbor period as the return trend correlation coefficient corresponding to the second nearest neighbor period.

[0350] The return trend correlation coefficient can be used to describe the correlation between the product's daily return trend data and the industry's daily return trend data. The higher the correlation, the larger the position of the sample industry in the sample resource management product.

[0351] Step 1914: obtain the product daily yield sequence and industry daily yield sequence in multiple third nearest neighbor time periods corresponding to each sample time node, and determine the slope coefficient and residual sum of squares corresponding to the corresponding third nearest neighbor time period based on the product daily yield sequence and the industry daily yield sequence.

[0352] The third nearest neighbor period may be a nearest neighbor period before the sample time node, for example, nearly one month, nearly two months, nearly three months, nearly six months, nearly one year, etc. before the sample time node.

[0353] The product daily yield sequence can represent the daily yield of the sample resource management product in the third nearest neighbor period, and the industry daily yield sequence can represent the daily yield of the sample industry in the third nearest neighbor period. The product daily yield sequence can be obtained based on the historical yield statistical data corresponding to the sample resource management product, and the daily yield of the sample resource management product in the third nearest neighbor period can be calculated based on the historical yield statistical data. The industry daily yield sequence can be obtained based on the industry yield statistical data corresponding to the sample industry, and the daily yield of the sample industry in the third nearest neighbor period can be calculated based on the industry yield statistical data.

[0354] Linear fitting is performed on the product daily yield sequence and the industry daily yield sequence corresponding to each third nearest neighbor period to obtain the linear fitting result. The linear fitting result can represent the relative change relationship between the product daily yield and the industry daily yield in the third nearest neighbor period. The slope coefficient corresponding to the linear fitting result can represent the relative change relationship between the product daily yield and the industry daily yield. A higher slope coefficient indicates that the product yield of the sample resource management product is more sensitive to the change of the industry yield of the sample industry, and the sample industry occupies a larger position in the sample resource management product. The residual sum of squares can represent the error between the true value corresponding to the product yield and the industry yield and the predicted value corresponding to the linear fitting result. The smaller the residual sum of squares, the closer the linear fitting result is to the true value corresponding to the product daily yield and the industry daily yield, the higher the accuracy of the linear fitting result in describing the relative change relationship between the product daily yield and the industry daily yield, and the higher the possibility that the sample resource management product contains the sample industry.

[0355] Step 1915: construct a sample yield relationship data sequence based on multiple yield differences corresponding to multiple first neighboring time periods corresponding to each sample time node, multiple yield trend correlation coefficients corresponding to multiple second neighboring time periods, and multiple slope coefficients and residual sum of squares corresponding to multiple third neighboring time periods.

[0356] The first adjacent time period may include a day and a week; the second adjacent time period and the second adjacent time period may include a recent 1 month, recent 2 months, recent 3 months, recent 6 months and recent 1 year.

[0357] Therefore, based on the multiple yield differences corresponding to the multiple first neighboring time periods, the multiple yield trend correlation coefficients corresponding to the multiple second neighboring time periods, and the multiple slope coefficients and residual sum of squares corresponding to the multiple third neighboring time periods, the sample yield relationship data corresponding to the sample time nodes are constructed. Therefore, the sample yield relationship data corresponding to each sample time node may include:

[0358] The difference between the daily return of the sample resource management product and the daily return of the industry;

[0359] The difference between the weekly return of the sample resource management product and the weekly return of the industry;

[0360] The correlation coefficient between the daily yield of sample resource management products and the daily yield of the industry in the past month;

[0361] The correlation coefficient between the daily yield of sample resource management products and the daily yield of the industry in the past two months;

[0362] The correlation coefficient between the daily yield of sample resource management products and the daily yield of the industry in the past three months;

[0363] The correlation coefficient between the daily yield of sample resource management products and the daily yield of the industry in the past six months;

[0364] The correlation coefficient between the daily yield of sample resource management products and the daily yield of the industry in the past year;

[0365] The linear regression slope coefficient of the daily returns of sample resource management products and the daily returns of the industry in the past month;

[0366] The sum of squares of the linear regression slopes of the daily returns of sample resource management products and the daily returns of the industry in the past month;

[0367] The linear regression slope coefficient of the daily yield of sample resource management products and the daily yield of the industry in the past two months;

[0368] The sum of squares of the linear regression slopes of the daily returns of sample resource management products and the daily returns of the industry in the past two months;

[0369] The linear regression slope coefficient of the daily yield of sample resource management products and the daily yield of the industry in the past three months;

[0370] The sum of squares of the linear regression slopes of the daily returns of sample resource management products and the daily returns of the industry in the past three months;

[0371] The linear regression slope coefficient of the daily yield of sample resource management products and the daily yield of the industry in the past six months;

[0372] The sum of squares of the linear regression slopes of the daily returns of sample resource management products and the daily returns of the industry in the past six months;

[0373] The linear regression slope coefficient of the daily return rate of sample resource management products and the daily return rate of the industry in the past year;

[0374] The sum of squares of the linear regression slopes of the daily returns of sample resource management products and the daily returns of the industry in the past year.

[0375] The sample-income relationship data corresponding to multiple sample time nodes are arranged in chronological order to construct a revenue relationship data sequence.

[0376] Step 1916: Construct known training sample data based on the sample return relationship data sequence and the industry position label sequence corresponding to multiple sample time nodes.

[0377] The sample return relationship data sequences corresponding to multiple sample time nodes are stored in correspondence with the industry position label sequences. Since the sample return relationship data sequences and the industry position label sequences are arranged in chronological order, the sample return relationship data in the sample return relationship data sequences correspond one-to-one to the industry position labels in the industry position label sequences.

[0378] Since the industry position label sequence is obtained based on the published industry position information statistical report, the training sample data constructed by the sample return relationship data sequence and the industry position label sequence is publicly known training sample data.

[0379] Step 1917: Perform linear interpolation processing based on multiple known training sample data to obtain first enhanced sample data.

[0380] Linear interpolation processing can perform interpolation numerical estimation based on the linear relationship between the two adjacent data points on the left and right of the point to be interpolated in the one-dimensional data sequence. For example, the sample time nodes corresponding to the sample income relationship data sequence are extracted according to the monthly cycle. In order to enhance the sample, the sample income relationship data extracted with a monthly cycle can be expanded to the sample income relationship data with a weekly cycle through linear difference processing. The sample income relationship data with a weekly cycle can be filled and interpolated based on the linear relationship corresponding to the monthly income relationship data.

[0381] Specifically, the industry position information corresponding to the interpolation time node in the two adjacent sample time nodes can be obtained according to the linear relationship between the two adjacent sample time nodes and the corresponding industry position labels in the known training sample data. The interpolation income relationship data corresponding to the interpolation time node is obtained from the historical income statistical data corresponding to the sample resource management product. The interpolation income relationship data corresponding to the interpolation time node is stored correspondingly to the industry position information to form the first enhanced sample data.

[0382] Step 1918: perform data splicing based on multiple known training sample data to obtain second enhanced sample data, and construct a training sample data set based on the first enhanced sample data, the second enhanced sample data and multiple known training sample data.

[0383] When performing data splicing, a connection variable can be added to the last bit of each sample profit relationship data in multiple sample profit relationship sequences in multiple known training sample data. The connection variable can be used to indicate whether the connection is the connection of the sample profit relationship data sequence corresponding to the sample resource management product. When the connection variable indicates that the connection is the connection of the sample profit relationship data in the sample profit relationship data sequence, the connection variable is 0; when the connection variable indicates that the connection is the connection of the sample profit relationship data sequence corresponding to the sample resource management product, the connection variable is 1. Similarly, a connection variable can be added to the last bit of each industry position label in multiple industry position label sequences in multiple known training sample data.

[0384] After adding the connection variable, multiple sample return relationship data sequences are spliced ​​head to tail, and multiple industry position label sequences are spliced ​​head to tail. The splicing order of multiple sample return relationship data sequences and multiple industry position label sequences is consistent based on the corresponding sample time nodes. The spliced ​​sample return relationship data sequences and industry position label sequences can construct the second enhanced sample data.

[0385] The training sample data set includes known training data and a training sample data set consisting of first enhanced sample data and second enhanced sample data.

[0386] Step 1919: Cross-train and verify the long short-term memory network model based on the training sample data set according to a predetermined division ratio to obtain a target neural network model.

[0387] First, the training sample data set can be divided into a training set and a validation set, for example, at a ratio of 8:2.

[0388] The sample data in the training set is evenly divided into k parts according to a predetermined division ratio, and the long short-term memory network model is trained k times. In each training, the sample data in the k-1 parts are used to train the long short-term memory network model, and the trained model is verified using the remaining one and the sample data in the verification set. When the verification result after k trainings meets the predetermined verification condition, for example, the average value of the prediction accuracy in the k verifications is greater than the predetermined threshold, the trained long short-term memory model network is used as the target neural network model.

[0389] The hyperparameter settings for training the LSTM network model may include: training sample size: 256; learning rate: 0.01; optimizer: Adam; loss function: mean absolute error (MAE); number of training cycles: 200

[0390] Step 1920: obtain the target benefit relationship data sequence corresponding to the target resource management product and the target industry, and use the target neural network model to perform position calculation on the target benefit relationship data sequence to obtain the industry position information occupied by the target industry in the target resource management product.

[0391] The target benefit relationship data sequence includes benefit relationship data collected at multiple time nodes corresponding to the target measurement moment. When obtaining benefit relationship data, the target benefit data corresponding to the multiple time nodes corresponding to the target resource management product at the target measurement moment, and the target industry benefit data corresponding to the target industry at multiple time nodes can be obtained. The benefit relationship data corresponding to multiple time nodes are determined based on the target benefit data and the target industry benefit data. The benefit relationship data can include the 17 data in step 1915. Therefore, the method for obtaining the benefit relationship data is the same as the method for obtaining multiple data in steps 1912-1914, which will not be repeated here.

[0392] By connecting the revenue relationship data corresponding to multiple time nodes in chronological order, a target revenue relationship data sequence can be obtained.

[0393] After obtaining the target benefit relationship data sequence corresponding to the target resource management product and the target industry, the target benefit relationship data sequence can be input into the target neural network model for position calculation, thereby obtaining the industry position information occupied by the target industry in the target resource management product at the target calculation moment.

[0394] In summary, the position calculation method of the resource management product of the embodiment of the present disclosure can be expressed as: Fig. 20 . First, the target neural network model is constructed through training sample data construction, training sample data enhancement, and long short-term memory network model training. Then, the target neural network model is used to calculate the industry position information for the target resource management product and the target industry. The position calculation method for the resource management product based on the embodiment of the present disclosure has an average measurement error of 1.74% for the position calculation of the target industry, which is significantly lower than the error of the calculation results obtained by other algorithms, greatly improving the accuracy of industry position calculation for resource management products.

[0395] Description of the apparatus and device of the present disclosure

[0396] It is to be understood that, although the steps in the above-mentioned flowcharts are sequentially displayed according to the characterization of arrows, these steps are not necessarily executed in sequence according to the order of arrow characterization. Unless there is a clear description in the present embodiment, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the above-mentioned flowcharts can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of the steps or stages in other steps.

[0397] It should be noted that in each specific embodiment of the present disclosure, when it comes to the need to perform relevant processing based on data related to the characteristics of the target object such as the target object attribute information or attribute information set, the permission or consent of the target object will be obtained first, and the collection, use and processing of these data will comply with the relevant laws, regulations and standards of the relevant region. In addition, when the embodiment of the present application needs to obtain the attribute information of the target object, the separate permission or separate consent of the target object will be obtained through a pop-up window or jump to a confirmation page. After clearly obtaining the separate permission or separate consent of the target object, the necessary target object-related data used to enable the normal operation of the embodiment of the present application will be obtained.

[0398] Fig.21 A schematic diagram of the structure of a storage position calculation device 2100 of a resource management product provided in an embodiment of the present disclosure. The device includes:

[0399] The first acquisition unit 2110 is used to acquire an industry position calculation request of a target resource management product, where the industry position calculation request includes target industry information and calculation time information;

[0400] The second acquisition unit 2120 is used to acquire first historical revenue statistics of the target resource management product and acquire second historical revenue statistics corresponding to the target industry information;

[0401] The determining unit 2130 is used to determine a plurality of measurement data collection time points based on the measurement time information, and determine a revenue relationship data sequence corresponding to the measurement time information according to the plurality of measurement data collection time points, the first historical revenue statistical data, and the second historical revenue statistical data;

[0402] The calculation unit 2140 is used to call the target neural network model to perform position calculation on the revenue relationship data sequence, and obtain the position information corresponding to the target industry in the target resource management product corresponding to the calculation time information.

[0403] Optionally, in one implementation, the determining unit 2130 is specifically configured to:

[0404] Extracting multiple node yield data from the first historical yield statistical data based on multiple measurement data collection time points, and extracting multiple industry yield data from the second historical yield statistical data;

[0405] The income relationship data corresponding to each measurement data collection time point is determined according to the node yield data and the corresponding industry yield data, and the income relationship data is sorted according to the time sequence of multiple measurement data collection time points to obtain the income relationship data sequence corresponding to the measurement time information.

[0406] Optionally, in one implementation, the determining unit 2130 is specifically configured to:

[0407] Calculate the difference between the node yield data and the industry yield data corresponding to each measurement data collection time point to obtain the first sub-yield relationship data;

[0408] Calculate the correlation coefficient between the first yield trend corresponding to the node yield data corresponding to each measured data collection time point and the second yield trend corresponding to the industry yield data to obtain the second sub-yield relationship data;

[0409] Perform correlation fitting between the node yield data and the industry yield data corresponding to each measurement data collection time point, and determine the fitting parameters obtained by fitting as the third sub-yield relationship data;

[0410] The revenue relationship data corresponding to each measurement data collection time point is determined according to the first sub-revenue relationship data, the second sub-revenue relationship data and the third sub-revenue relationship data.

[0411] Optionally, in one implementation, the determining unit 2130 is specifically configured to:

[0412] Obtain product yield values ​​and industry yield values ​​in multiple first nearest neighbor time periods corresponding to each measured data collection time point;

[0413] Calculate the difference between the product yield value and the industry yield value corresponding to each first neighbor time period, obtain multiple yield difference values ​​corresponding to multiple first neighbor time periods, and determine the first sub-yield relationship data based on the multiple yield difference values.

[0414] Optionally, in one implementation, the determining unit 2130 is specifically configured to:

[0415] Obtain product daily return trend data and industry daily return trend data in multiple second nearest neighbor time periods corresponding to each measured data collection time point;

[0416] Calculate the correlation coefficient between the product daily return trend data and the industry daily return trend data corresponding to each second nearest neighbor time period, obtain multiple correlation coefficients corresponding to multiple second nearest neighbor time periods, and determine the second sub-return relationship data based on the multiple correlation coefficients.

[0417] Optionally, in one implementation, the determining unit 2130 is specifically configured to:

[0418] Obtain product daily yield sequence and industry yield sequence in multiple third nearest neighbor time periods corresponding to each measured data collection time point;

[0419] Perform linear fitting on the product daily yield sequence and the industry yield sequence to obtain the first linear fitting result;

[0420] Determine a first fitting parameter corresponding to each third nearest neighbor time period based on a slope coefficient of the first linear fitting result, and determine a second fitting parameter corresponding to each third nearest neighbor time period based on a residual square sum of the first linear fitting result;

[0421] The third sub-income relationship data is determined according to the first fitting parameter and the second fitting parameter.

[0422] Optionally, in one embodiment, the training process of the target neural network model includes:

[0423] Acquire multiple training sample data, the training sample data including a sample income relationship data sequence between a sample resource management product and a sample industry and a corresponding industry position label sequence, the sample income relationship data sequence being a sequence of sample income relationship data corresponding to multiple sample time nodes in chronological order;

[0424] The sample return relationship data sequence in multiple training sample data is used as the model input, and the corresponding industry position label sequence is used as the model training label training preset recurrent neural network model to obtain the trained target neural network model.

[0425] Optionally, in one implementation, a plurality of training sample data are obtained, the training sample data comprising a sample revenue relationship data sequence between a sample resource management product and a sample industry and a corresponding industry position label sequence, the sample revenue relationship data sequence being a sequence of sample revenue relationship data corresponding to a plurality of sample time nodes in chronological order, including:

[0426] Obtain the third historical income statistics and historical industry position information statistics of multiple sample resource management products within a preset period, and obtain the industry income statistics corresponding to resource management products in multiple industries within a preset period;

[0427] Determine multiple sample time nodes within a preset period, and determine sample industries among multiple industries;

[0428] Extracting a sample income relationship data sequence corresponding to each sample resource management product from the third historical income statistical data and the industry income statistical data based on multiple sample time nodes and sample industries;

[0429] Determine the industry position label sequence corresponding to each sample return relationship data sequence based on multiple sample time nodes and historical industry position information statistical data;

[0430] A plurality of training sample data are determined based on a plurality of sample resource management products and a sample industry's sample return relationship data sequence and a corresponding industry position label sequence.

[0431] Optionally, in one embodiment, before determining a plurality of training sample data according to a plurality of sample resource management products and sample industry sample revenue relationship data sequences and corresponding industry position label sequences, the method further includes:

[0432] Perform linear interpolation processing on the sample income relationship data sequence corresponding to each sample resource management product and the corresponding industry position label sequence to obtain the first sub-enhanced sample data;

[0433] Data splicing is performed on the sample income relationship data sequences corresponding to multiple sample resource management products and the corresponding industry position label sequences to obtain the second sub-enhanced sample data;

[0434] Determine enhanced sample data according to the first sub-enhanced sample data and the second sub-enhanced sample data;

[0435] A plurality of training sample data are determined according to a plurality of sample resource management products and a sample industry sample income relationship data sequence and a corresponding industry position label sequence, including:

[0436] Multiple training sample data are determined based on enhanced sample data, sample return relationship data sequences of multiple sample resource management products and sample industries, and corresponding industry position label sequences.

[0437] Optionally, in one implementation, a linear interpolation process is performed on the sample revenue relationship data sequence and the corresponding industry position label sequence corresponding to each sample resource management product to obtain first sub-enhanced sample data, including:

[0438] Obtain the first sampling time and first industry position information corresponding to each sample income relationship data in the sample income relationship data sequence corresponding to each sample resource management product;

[0439] Performing a linear fit on the first sampling time and the first industry position information to obtain a second linear fit result;

[0440] Perform interpolation processing based on the second linear fitting result to obtain multiple second sampling times and corresponding second industry position information;

[0441] The enhanced profit relationship data corresponding to each second sampling time is obtained, and the first sub-enhanced sample data is determined according to the enhanced profit relationship data and the second industry position information.

[0442] Optionally, in one implementation, data splicing is performed on sample revenue relationship data sequences corresponding to multiple sample resource management products and corresponding industry position label sequences to obtain second sub-enhanced sample data, including:

[0443] Perform feature concatenation on the sample revenue relationship data sequences corresponding to multiple sample resource management products to obtain enhanced sample features;

[0444] The industry position label sequences corresponding to multiple sample resource management products are sequentially spliced ​​to obtain enhanced sample labels, and the second sub-enhanced sample data is determined according to the enhanced sample features and the enhanced sample labels.

[0445] Optionally, in one embodiment, a sample return relationship data sequence in a plurality of training sample data is used as a model input, and a corresponding industry position label sequence is used as a model training label training preset recurrent neural network model to obtain a trained target neural network model, including:

[0446] Dividing the plurality of training sample data into a first training set and a first validation set;

[0447] Randomly extract a preset proportion of target training sample data from the first training set, and construct a second validation set with the extracted target training sample data and the first validation set, and construct a second training set with the training sample data other than the target training sample data in the first training set;

[0448] A preset long short-term memory network model is trained based on the sample return relationship data sequence in the second training set as the model input and the corresponding industry position label sequence as the model training label;

[0449] The trained long short-term memory network model is verified according to the second verification set, and when the verification result meets the preset verification condition, the trained long short-term memory network model is determined as the target neural network model.

[0450] In the embodiments of the present application, the term "module" or "unit" refers to a computer program or a part of a computer program with a predetermined function, and works together with other related parts to achieve a predetermined goal, and can be implemented in whole or in part by using software, hardware (such as processing circuits or memories) or a combination thereof. Similarly, a processor (or multiple processors or memories) can be used to implement one or more modules or units. In addition, each module or unit can be part of an overall module or unit that includes the function of the module or unit.

[0451] Reference Fig. 22 , Fig. 22The structural block diagram of the terminal 140 for implementing the method for calculating the position of the resource management product of the embodiment of the present disclosure is as follows: the terminal 140 includes: a radio frequency (RF) circuit 2210, a memory 2215, an input unit 2230, a display unit 2240, a sensor 2250, an audio circuit 2260, a wireless fidelity (WiFi) module 2270, a processor 2280, and a power supply 2290. Those skilled in the art can understand that Fig. 22 The structure of the terminal 140 shown does not constitute a limitation on a mobile phone or a computer, and may include more or less components than shown in the figure, or combine certain components, or arrange the components differently.

[0452] The RF circuit 2210 may be used for receiving and sending signals during information transmission or calls. In particular, after receiving the downlink information from the base station, it is sent to the processor 2280 for processing; in addition, the designed uplink data is sent to the base station.

[0453] The memory 2215 may be used to store software programs and modules. The processor 2280 executes various functional applications and document editing of the terminal by running the software programs and modules stored in the memory 2215 .

[0454] The input unit 2230 may be used to receive input digital or character information and generate key signal input related to the terminal's settings and function control. Specifically, the input unit 2230 may include a touch panel 2231 and other input devices 2232 .

[0455] The display unit 2240 may be used to display input information or provided information and various menus of the terminal. The display unit 2240 may include a display panel 2241.

[0456] The audio circuit 2260, the speaker 2261, and the microphone 2262 may provide an audio interface.

[0457] In this embodiment, the processor 2280 included in the terminal 140 can execute the position calculation method of the resource management product in the previous embodiment.

[0458] The terminal 140 of the embodiment of the present disclosure includes but is not limited to mobile phones, computers, intelligent voice interaction devices, smart home appliances, vehicle-mounted terminals, aircraft, etc.

[0459] Fig.23A block diagram of the structure of a portion of a server 110 for implementing the method for calculating the position of a resource management product according to an embodiment of the present disclosure. The server 110 may have relatively large differences due to different configurations or performances, and may include one or more central processing units (CPUs) 2322 (e.g., one or more processors) and a storage device 2332, and one or more storage media 2330 (e.g., one or more mass storage devices) storing application programs 2342 or data 2344. Among them, the storage device 2332 and the storage medium 2330 may be temporary storage or permanent storage. The program stored in the storage medium 2330 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations on the server 110. Furthermore, the central processing unit 2322 may be configured to communicate with the storage medium 2330 to execute a series of instruction operations in the storage medium 2330 on the server 110.

[0460] The server 110 may also include one or more power supplies 2326, one or more wired or wireless network interfaces 2350, one or more input and output interfaces 2358, and / or one or more operating systems 2341, such as Windows ServerTM, Mac OS XTM, UnixTM, LinuxTM, FreeBSDTM, etc.

[0461] The central processor 2322 in the server 110 can be used to execute the position calculation method of the resource management product of the embodiment of the present disclosure.

[0462] The disclosed embodiment also provides a storage medium for storing program code, and the program code is used to execute the position calculation method of the resource management product of each of the aforementioned embodiments.

[0463] The embodiment of the present disclosure also provides a computer program product, which includes a computer program. A processor of an electronic device reads and executes the computer program, so that the electronic device executes the above-mentioned method for calculating the position of the resource management product.

[0464] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present disclosure described herein can, for example, be implemented in an order other than those illustrated or described herein. In addition, the terms "comprises" and "comprising" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0465] It should be understood that in the present disclosure, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0466] It should be understood that in the description of the embodiments of the present disclosure, the meaning of multiple (or multiple items) is more than two, greater than, less than, exceed, etc. are understood to not include the number, and above, below, within, etc. are understood to include the number.

[0467] In the several embodiments provided in the present disclosure, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0468] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0469] In addition, each functional unit in each embodiment of the present disclosure may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0470] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a storage medium. Based on this understanding, the technical solution of the present disclosure is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the various embodiments of the present disclosure. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), disk or optical disk and other media that can store program codes.

[0471] It should also be understood that the various implementations provided in the embodiments of the present disclosure can be combined arbitrarily to achieve different technical effects.

[0472] The above is a specific description of the implementation methods of the present disclosure, but the present disclosure is not limited to the above implementation methods. Technical personnel familiar with the art can also make various equivalent modifications or substitutions without violating the spirit of the present disclosure. These equivalent modifications or substitutions are all included in the scope defined by the claims of the present disclosure.

Claims

1. A method for calculating the position of a resource management product, characterized in that: The method comprises: Obtain an industry position calculation request for a target resource management product, wherein the industry position calculation request includes target industry information and calculation time information; Acquire first historical revenue statistics of the target resource management product and acquire second historical revenue statistics corresponding to the target industry information; Determine a plurality of measurement data collection time points based on the measurement time information, and determine a revenue relationship data sequence corresponding to the measurement time information according to the plurality of measurement data collection time points, the first historical revenue statistical data, and the second historical revenue statistical data; The target neural network model is called to perform position calculation on the revenue relationship data sequence to obtain the position information corresponding to the target industry in the target resource management product corresponding to the calculation time information.

2. The method according to claim 1, characterized in that The determining, according to the multiple measurement data collection time points, the first historical revenue statistical data, and the second historical revenue statistical data, a revenue relationship data sequence corresponding to the measurement time information includes: Extracting a plurality of node yield data from the first historical yield statistical data based on the plurality of measured data collection time points, and extracting a plurality of industry yield data from the second historical yield statistical data; The profit relationship data corresponding to each of the measurement data collection time points is determined according to the node yield data and the corresponding industry yield data, and the profit relationship data is sorted according to the time sequence of the multiple measurement data collection time points to obtain a profit relationship data sequence corresponding to the measurement time information.

3. The method according to claim 2, characterized in that The determining of the profit relationship data corresponding to each of the measured data collection time points according to the node profit rate data and the corresponding industry profit rate data includes: Calculate the difference between the node yield data and the industry yield data corresponding to each of the measured data collection time points to obtain the first sub-yield relationship data; Calculate the correlation coefficient between the first yield trend corresponding to the node yield data corresponding to each of the measured data collection time points and the second yield trend corresponding to the industry yield data to obtain the second sub-yield relationship data; Perform correlation fitting between the node yield data and the industry yield data corresponding to each of the measured data collection time points, and determine the fitting parameters obtained by fitting as the third sub-yield relationship data; The revenue relationship data corresponding to each of the measurement data collection time points is determined according to the first sub-revenue relationship data, the second sub-revenue relationship data and the third sub-revenue relationship data.

4. The method according to claim 3, characterized in that The calculation of the difference between the node yield data and the industry yield data corresponding to each of the measured data collection time points to obtain the first sub-yield relationship data includes: Obtain product yield values ​​and industry yield values ​​in multiple first nearest neighbor time periods corresponding to each of the measured data collection time points; Calculate the difference between the product yield value and the industry yield value corresponding to each of the first neighboring time periods to obtain multiple yield difference values ​​corresponding to the multiple first neighboring time periods, and determine the first sub-yield relationship data based on the multiple yield difference values.

5. The method according to claim 3, characterized in that: The calculation of the correlation coefficient between the first yield trend corresponding to the node yield data corresponding to each of the measured data collection time points and the second yield trend corresponding to the industry yield data to obtain the second sub-yield relationship data includes: Obtain product daily return trend data and industry daily return trend data in multiple second nearest neighbor time periods corresponding to each of the measured data collection time points; Calculate the correlation coefficient between the product daily return trend data and the industry daily return trend data corresponding to each of the second neighboring time periods, obtain multiple correlation coefficients corresponding to the multiple second neighboring time periods, and determine the second sub-return relationship data based on the multiple correlation coefficients.

6. The method according to claim 3, characterized in that The correlation fitting is performed between the node yield data corresponding to each of the measured data collection time points and the industry yield data, and the fitting parameters obtained by the fitting are determined as the third sub-yield relationship data, including: Obtain product daily yield sequences and industry yield sequences in multiple third nearest neighbor time periods corresponding to each of the measured data collection time points; Performing linear fitting on the product daily yield sequence and the industry yield sequence to obtain a first linear fitting result; Determine a first fitting parameter corresponding to each of the third nearest neighbor time periods based on a slope coefficient of the first linear fitting result, and determine a second fitting parameter corresponding to each of the third nearest neighbor time periods based on a residual square sum of the first linear fitting result; The third sub-income relationship data is determined according to the first fitting parameter and the second fitting parameter.

7. The method according to claim 1, characterized in that The training process of the target neural network model includes: Acquire multiple training sample data, wherein the training sample data includes a sample income relationship data sequence between a sample resource management product and a sample industry and a corresponding industry position label sequence, wherein the sample income relationship data sequence is a sequence of sample income relationship data corresponding to multiple sample time nodes in chronological order; The sample return relationship data sequence in the multiple training sample data is used as the model input, and the corresponding industry position label sequence is used as the model training label training preset recurrent neural network model to obtain the trained target neural network model.

8. The method according to claim 7, characterized in that The obtaining of a plurality of training sample data includes a sample revenue relationship data sequence between a sample resource management product and a sample industry and a corresponding industry position label sequence, wherein the sample revenue relationship data sequence is a sequence of sample revenue relationship data corresponding to a plurality of sample time nodes in chronological order, including: Obtain third historical revenue statistics and historical industry position information statistics of multiple sample resource management products within a preset period of time, and obtain industry revenue statistics corresponding to resource management products of multiple industries within the preset period of time; Determining a plurality of sample time nodes within the preset period, and determining a sample industry among the plurality of industries; Extracting a sample revenue relationship data sequence corresponding to each of the sample resource management products from the third historical revenue statistical data and the industry revenue statistical data based on the multiple sample time nodes and the sample industry; Determine the industry position label sequence corresponding to each of the sample return relationship data sequences based on the multiple sample time nodes and the historical industry position information statistical data; A plurality of training sample data are determined based on the sample revenue relationship data sequences of the plurality of sample resource management products and sample industries and the corresponding industry position label sequences.

9. The method according to claim 8, characterized in that Before determining a plurality of training sample data according to the plurality of sample resource management products and sample industry sample revenue relationship data sequences and the corresponding industry position label sequences, the method further includes: Performing linear interpolation processing on the sample income relationship data sequence and the corresponding industry position label sequence corresponding to each of the sample resource management products to obtain first sub-enhanced sample data; Performing data splicing on the sample income relationship data sequences and the corresponding industry position label sequences corresponding to the multiple sample resource management products to obtain second sub-enhanced sample data; Determine enhanced sample data according to the first sub-enhanced sample data and the second sub-enhanced sample data; The determining of a plurality of training sample data according to the plurality of sample resource management products and sample industry sample revenue relationship data sequences and the corresponding industry position label sequences comprises: A plurality of training sample data are determined according to the enhanced sample data, the sample revenue relationship data sequences of the plurality of sample resource management products and sample industries, and the corresponding industry position label sequences.

10. The method according to claim 9, characterized in that The linear interpolation process is performed on the sample revenue relationship data sequence and the corresponding industry position label sequence corresponding to each of the sample resource management products to obtain the first sub-enhanced sample data, including: Obtaining the first sampling time and the first industry position information corresponding to each sample benefit relationship data in the sample benefit relationship data sequence corresponding to each of the sample resource management products; Performing a linear fit on the first sampling time and the first industry position information to obtain a second linear fit result; Perform interpolation processing based on the second linear fitting result to obtain multiple second sampling times and corresponding second industry position information; Obtain enhanced revenue relationship data corresponding to each of the second sampling times, and determine first sub-enhanced sample data based on the enhanced revenue relationship data and the second industry position information.

11. The method according to claim 9, characterized in that The step of performing data splicing on the sample revenue relationship data sequences and the corresponding industry position label sequences corresponding to the multiple sample resource management products to obtain the second sub-enhanced sample data includes: Perform feature concatenation on the sample revenue relationship data sequences corresponding to multiple sample resource management products to obtain enhanced sample features; The industry position label sequences corresponding to the multiple sample resource management products are sequentially spliced ​​to obtain enhanced sample labels, and second sub-enhanced sample data is determined according to the enhanced sample features and the enhanced sample labels.

12. The method according to claim 7, characterized in that The method of using the sample return relationship data sequence in the plurality of training sample data as model input and using the corresponding industry position label sequence as model training label training preset recurrent neural network model to obtain the trained target neural network model includes: Dividing the plurality of training sample data into a first training set and a first validation set; Randomly extract a preset proportion of target training sample data from the first training set, and construct a second validation set with the extracted target training sample data and the first validation set, and construct a second training set with the training sample data other than the target training sample data in the first training set; A preset long short-term memory network model is trained based on the sample return relationship data sequence in the second training set as a model input and the corresponding industry position label sequence as a model training label; The trained long short-term memory network model is verified according to the second verification set, and when the verification result meets the preset verification condition, the trained long short-term memory network model is determined as the target neural network model.

13. A storage position calculation device for a resource management product, characterized in that: The device comprises: A first acquisition unit is used to acquire an industry position calculation request of a target resource management product, wherein the industry position calculation request includes target industry information and calculation time information; A second acquisition unit is used to acquire first historical revenue statistics of the target resource management product and to acquire second historical revenue statistics corresponding to the target industry information; a determining unit, configured to determine a plurality of measurement data collection time points based on the measurement time information, and determine a revenue relationship data sequence corresponding to the measurement time information according to the plurality of measurement data collection time points, the first historical revenue statistical data, and the second historical revenue statistical data; The calculation unit is used to call the target neural network model to perform position calculation on the revenue relationship data sequence to obtain the position information corresponding to the target industry in the target resource management product corresponding to the calculation time information.

14. A storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method for calculating the position of a resource management product according to any one of claims 1 to 12 is implemented.

15. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the position calculation method of the resource management product described in any one of claims 1 to 12 is implemented.

16. A computer program product, comprising a computer program, wherein the computer program is read and executed by a processor of an electronic device, so that the electronic device executes the position calculation method of the resource management product described in any one of claims 1 to 12.

Images