Data processing method, device and electronic equipment

By extracting word vector groups and dynamic indicators of objects and categories, and using neural network models to determine dynamic indicators of stock concepts, the problem of automatic identification of multiple categories and data influence in the existing technology is solved, and the accuracy and timeliness of data mining are improved.

CN112750042BActive Publication Date: 2024-11-19TENCENT TECHNOLOGY (SHENZHEN) CO LTD
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
CN201911044056.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2019-10-30
Publication Date
2024-11-19
Estimated Expiration
2039-12-10

AI Technical Summary

Technical Problem

In the field of data mining, especially in the process of stock index compilation, it is difficult for the existing technology to automatically identify multiple categories involved in individual objects and their impact on categories, and manual labeling methods cannot identify emerging business or stock concepts in a timely manner. This results in complex and inaccurate data processing.

Method used

By obtaining the text data of the object, extracting multiple word vector groups of the corresponding relationship between the object and the category, and combining the object's dynamic indicators, the preset neural network model is used to determine the dynamic indicators of the category. The specific steps include calculating the weight of the object to the target category and dynamic indicators to determine the dynamic changes of the category.

Benefits of technology

It realizes automatic identification of categories involved in objects and their impact on category data, improves the accuracy of data mining, and timely identifys new categories, avoiding the problems of untimely and inaccurate information marked manually.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN112750042B_ABST
    Figure CN112750042B_ABST
Patent Text Reader

Abstract

The present disclosure provides a data processing method, device and electronic device. The data processing method includes: obtaining text data including an object; extracting multiple word vector groups including the correspondence between the object and the category from the text data; obtaining the dynamic index of the object; and determining the dynamic index of the category according to the multiple word vector groups and the dynamic index of the object. The data processing method provided by the present disclosure can automatically obtain the dynamic index of multiple categories according to the text data related to the object and the dynamic index of the object.
Need to check novelty before this filing date? Find Prior Art

Description

Background Art

[0002] In the field of data mining, it is often necessary to obtain data corresponding to multiple categories based on the data of a large number of individual objects, such as obtaining the stock index of a certain stock concept based on the stock price change rate of individual stocks. When the individual object clearly belongs to a certain category, the data processing process is relatively simple, but when the individual object belongs to multiple categories at the same time, how to determine the impact of the individual object's data on the category data often becomes very complicated.

[0003] In related technologies, it is often necessary to manually label which categories an individual object is involved in, or even label the proportion of an individual object in each category. However, in some complex scenarios, such as the compilation of stock indexes, listed companies often involve multiple businesses, and their data can affect multiple stock concepts. Manual labeling methods often have difficulty dealing with the complex business areas of listed companies, and are unable to promptly identify emerging businesses and new stock concepts.

[0004] Therefore, there is a need for a data mining technology solution that can automatically identify the categories involved in individual objects and automatically determine the impact of data changes of individual objects on data changes of categories.

[0005] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present disclosure, and therefore may include information that does not constitute the prior art known to ordinary technicians in the field. Summary of the invention

[0006] The purpose of the present disclosure is to provide a data processing method, device and electronic device, which are used to overcome, at least to a certain extent, the problem that it is difficult to accurately determine the relationship between individual object data and category data in the data mining process due to the limitations and defects of related technologies.

[0007] According to a first aspect of an embodiment of the present disclosure, a data processing method is provided, comprising: acquiring text data including an object; extracting a plurality of word vector groups including correspondences between objects and categories from the text data; acquiring a dynamic index of the object; and determining a dynamic index of the category based on the plurality of word vector groups and the dynamic index of the object.

[0008] In an exemplary embodiment of the present disclosure, determining the dynamic index of the category according to the multiple word vector groups and the dynamic index of the object includes:

[0009] Determining weights of multiple objects for a target category according to the multiple word vector groups;

[0010] Determining the product of the weight of each object for the target category and the dynamic index of the object;

[0011] Taking the sum of multiple products corresponding to multiple objects as a numerator;

[0012] Taking the sum of the weights of the multiple objects to the target category as the denominator;

[0013] The dynamic index corresponding to the target category is determined according to the ratio of the numerator to the denominator.

[0014] In an exemplary embodiment of the present disclosure, determining the weights of multiple objects to a target category according to the multiple word vector groups includes:

[0015] Inputting the multiple word vector groups into a preset neural network model to obtain multiple output values;

[0016] Determine m output values ​​corresponding to the word vector group including both the target object and the target category from the multiple output values, where m≥1;

[0017] Determine n output values ​​corresponding to the word vector group including the target object from the multiple output values, where n≥1;

[0018] The weight of the target object to the target category is determined according to the ratio of the sum of the m output values ​​to the sum of the n output values.

[0019] In an exemplary embodiment of the present disclosure, the loss function of the preset neural network model includes:

[0020]

[0021] Among them, L is the loss function, j is the serial number of the target category, i1 and i2 are the serial numbers of the target objects, w is the weight of the target object to the target category, and x is the dynamic indicator of the target object.

[0022] In an exemplary embodiment of the present disclosure, the training process of the preset neural network model includes:

[0023] Initializing the network model parameters of the preset neural network model, or setting the network model parameters of the previous preset time length as the current network model parameters of the preset neural network model;

[0024] Acquire a training data set including the text data, and extract the text data to obtain a plurality of word vector groups;

[0025] Inputting the word vector group corresponding to the preset time length into the preset neural network model, so that the preset neural network model adjusts the network model parameters to minimize the value of the loss function;

[0026] The preset neural network model is trained using word vector groups corresponding to multiple preset time lengths, and the final network model parameters of the preset neural network model are determined after the training is completed.

[0027] In an exemplary embodiment of the present disclosure, the object is a name of a listed company, the category is a stock concept, and the dynamic indicator is a stock price change rate.

[0028] In an exemplary embodiment of the present disclosure, the word vector group includes the corresponding relationship between the name of the listed company, the behavior and the stock concept.

[0029] According to a second aspect of an embodiment of the present disclosure, there is provided a data processing device, including:

[0030] A text data acquisition module, configured to acquire text data including an object;

[0031] A relational data extraction module, configured to extract a plurality of word vector groups including corresponding relationships between objects and categories from the text data;

[0032] An object dynamic index acquisition module, configured to acquire the dynamic index of the object;

[0033] The category dynamic index determination module is configured to determine the dynamic index of the category based on the multiple word vector groups and the dynamic index of the object.

[0034] According to a third aspect of the present disclosure, there is provided an electronic device, comprising: a memory; and a processor coupled to the memory, wherein the processor is configured to execute any one of the methods described above based on instructions stored in the memory.

[0035] According to a fourth aspect of the present disclosure, a computer-readable storage medium is provided, on which a program is stored, and when the program is executed by a processor, the data processing method as described in any one of the above items is implemented.

[0036] The disclosed embodiments extract the correspondence between objects and categories through a large amount of text data, and input the correspondence data and the dynamic indicators of the objects into a trained data processing model, so as to accurately determine the relationship between objects and categories, the relationship between the data of objects and the data of categories, timely identify new categories, more accurately determine the data corresponding to categories, and use machine learning methods to improve the accuracy of data mining.

[0037] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The accompanying drawings herein are incorporated into the specification and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the specification are used to explain the principles of the present disclosure. Obviously, the accompanying drawings described below are only some embodiments of the present disclosure, and for ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without creative work.

[0039] Figure 1 is a flow chart of a data processing method in an exemplary embodiment of the present disclosure.

[0040] Figure 2 It is a sub-step flow chart of step S4 in the exemplary embodiment of the present disclosure.

[0041] Figure 3 It is a sub-step flowchart of step S41 in the exemplary embodiment of the present disclosure.

[0042] Figure 4 It is a flow chart of the training process of the neural network model in the exemplary embodiment of the present disclosure.

[0043] Figure 5 It is a schematic diagram of an application scenario of an embodiment of the present disclosure.

[0044] Figure 6 is a block diagram of a data processing device in an exemplary embodiment of the present disclosure.

[0045] Figure 7 is a block diagram of an electronic device in an exemplary embodiment of the present disclosure. DETAILED DESCRIPTION

[0046] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as being limited to the examples set forth herein; on the contrary, these embodiments are provided so that the present disclosure will be more comprehensive and complete, and the concepts of the example embodiments are fully conveyed to those skilled in the art. The described features, structures, or characteristics may be combined in one or more embodiments in any suitable manner. In the following description, many specific details are provided to provide a full understanding of the embodiments of the present disclosure. However, those skilled in the art will appreciate that the technical solutions of the present disclosure may be practiced while omitting one or more of the specific details, or other methods, components, devices, steps, etc. may be adopted. In other cases, known technical solutions are not shown or described in detail to avoid obscuring various aspects of the present disclosure.

[0047] In addition, the accompanying drawings are only schematic illustrations of the present disclosure, and the same reference numerals in the drawings represent the same or similar parts, so their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities, which do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software form, or in one or more hardware modules or integrated circuits, or in different network models and / or processor devices and / or microcontroller devices.

[0048] The exemplary embodiments of the present disclosure are described in detail below with reference to the accompanying drawings.

[0049] Figure 1 The flowchart of the data processing method in the exemplary embodiment of the present disclosure is schematically shown. Figure 1 , the data processing method 100 may include:

[0050] Step S1, obtaining text data including an object;

[0051] Step S2, extracting a plurality of word vector groups including corresponding relationships between objects and categories from the text data;

[0052] Step S3, obtaining dynamic indicators of the object;

[0053] Step S4, determining the dynamic index of the category according to the multiple word vector groups and the dynamic index of the object.

[0054] The disclosed embodiments extract the correspondence between objects and categories through a large amount of text data, and input the correspondence data and the dynamic indicators of the objects into a trained data processing model, so as to accurately determine the relationship between objects and categories, the relationship between the data of objects and the data of categories, timely identify new categories, more accurately determine the data corresponding to categories, and use machine learning methods to improve the accuracy of data mining.

[0055] In step S1, text data including an object is acquired.

[0056] Text data can be obtained from a variety of sources depending on the type of object. For example, when the object is the name of a listed company, text data related to the name of the listed company can be obtained from sources such as WeChat public accounts and self-selected stock financial information web pages. The method of obtaining text data can be, for example, crawling through a crawler. In some embodiments, it is possible to specify which objects (listed companies) to obtain text data related to based on data such as a list of listed companies. Of course, the sources of the above text data are only examples, and those skilled in the art can choose more sources to improve the accuracy of data recognition.

[0057] In step S2, a plurality of word vector groups including correspondences between objects and categories are extracted from the text data.

[0058] In the disclosed embodiments, data extraction can be performed on text data through a graph network model. A graph network model (GN) is a set of functions organized in a topological space according to a graph structure for relational reasoning. In deep learning theory, it is a generalization of a graph neural network model (GNN) and a probabilistic graphical model (PGM). The graph network model is composed of a graph network model block (GN block), has a flexible topological structure, and can be specialized into various connectionist models, including a feedforward neural network model (Feedforward Neural Network), a recursive neural network model (Recursive Neural Network), and the like. A more general graph network model is suitable for processing data with a graph structure, such as a knowledge graph, a social network model, a molecular network model, and the like. In the disclosed embodiments, the graph network models used include, but are not limited to, probabilistic graph models, graph convolution models, deep probabilistic graph models, hidden Markov models, and the like.

[0059] In one embodiment of the present disclosure, knowledge graph extraction can be implemented by a neural network model trained based on a deep learning mechanism. The structure of the neural network model can include multiple types, such as a feedforward neural network model, a recursive neural network model, etc. The above-mentioned graph network model can be connected with a neural network model of dynamic indicators of the recognition category to form a data processing model.

[0060] The basis for extracting word vector groups by the graph network model can be determined according to the type of object being processed. For example, when the data processing task is stock index compilation, the object is the name of a listed company, and the category is a stock concept, the text content can be extracted according to the "financial knowledge graph" as a ternary word vector group in the form of (company i, behavior a, concept j) such as (xx company, investment, virtual asset concept).

[0061] It is worth mentioning that for the stock index compilation task, the word vector group can be in the form of the triple shown above. Because the simple correspondence between the company and the concept is not enough to explain which aspect of the concept the company affects, for example, the company that invests in concept A, the company that develops concept A, and the listed company that operates concept A have different contributions to the stock index changes of concept A. Therefore, in the embodiment of the present disclosure, the reference indicator of "action" of listed companies is added to more accurately determine the degree of influence of each listed company on the data of a stock concept. In other data processing tasks, those skilled in the art can determine other correspondences besides the correspondence between objects and categories to improve the accuracy of judgment, and the present disclosure does not impose any special restrictions on this.

[0062] In step S3, the dynamic index of the object is obtained.

[0063] In one embodiment, the dynamic indicator is, for example, the price change rate of a stock.

[0064] We can use the day as the time period, obtain the text data of the day, and obtain the stock price change rate of all listed companies on the day from the public market information, and calculate the stock price change rate x of company i using the following formula (1):

[0065]

[0066] Among them, P t is the closing price of company i’s stock on day t, P t-1 is the closing price of company i’s stock on day t-1.

[0067] In step S4, the dynamic index of the category is determined according to the multiple word vector groups and the dynamic index of the object.

[0068] In the disclosed embodiment, the above data processing task is completed by a trained neural network model, and dynamic indicators of multiple categories are determined according to dynamic indicators of multiple word vector groups and objects.

[0069] Figure 2 It is a sub-step flow chart of step S4.

[0070] refer to Figure 2 , step S4 may include:

[0071] Step S41, determining weights of multiple objects to a target category according to the multiple word vector groups;

[0072] Step S42, determining the product of the weight of each object to the target category and the dynamic index of the object;

[0073] Step S43, taking the sum of the multiple products corresponding to the multiple objects as the numerator;

[0074] Step S44, taking the sum of the weights of the multiple objects to the target category as the denominator;

[0075] Step S45: determining the dynamic index corresponding to the target category according to the ratio of the numerator to the denominator.

[0076] Taking the stock index compilation task as an example, for a trained preset neural network model (network model structure is f, network model parameters are w), all companies i, behaviors a, and concepts j can be initialized as word vectors c i 、v a 、h j , the triple word vector group is described as (c i ,v a ,h j ) in the form of, and multiple triple word vector groups are input into the preset neural network model. In the embodiment of the present disclosure, the triple word vector groups can be input into the preset neural network model as three word vectors at the same time, or the triple word vector groups can be connected into one word vector and input into the preset neural network model, and the present disclosure is not limited to this.

[0077] The main process of determining concept j by the preset neural network model can be described by formula (2):

[0078]

[0079] Where j is the serial number of concept j, i is the serial number of company i, and w ij is the weight of company i on concept j, (i,,j) refers to the word vector group involving both company i and concept j, z j is the rate of change of the stock index corresponding to concept j.

[0080] The weight of company i on concept j refers to the proportion of company i's business areas involving concept j, w ij It can be calculated first by formula (3):

[0081]

[0082] Among them, f is the neural network model, and (i,,) refers to the entire word vector group involving company i.

[0083] The meaning of formula (3) can be expressed by Figure 3 The flowchart shown is represented.

[0084] refer to Figure 3 , step S41 may include:

[0085] Step S411, inputting the multiple word vector groups into a preset neural network model to obtain multiple output values;

[0086] Step S412, determining m output values ​​corresponding to the word vector group including both the target object and the target category from the multiple output values, m≥1;

[0087] Step S413, determining n output values ​​corresponding to the word vector group including the target object from the multiple output values, where n≥1;

[0088] Step S414: determining the weight of the target object to the target category according to the ratio of the sum of the m output values ​​to the sum of the n output values.

[0089] In formula (3), Σ (i,,j) f(c i ,v a ,h j |W) corresponds to the sum of the m output values ​​in the above text description, Σ (i,,) f(c i ,v a ,h j |W) corresponds to the sum of the n output values ​​in the above text description.

[0090] The structure of the neural network model can be various, for example, it can at least include i 、v a 、h j Three layers of processing were performed.

[0091] Figure 4 It is a flow chart of the training process of the neural network model.

[0092] refer to Figure 4 , the training process of the neural network model can include:

[0093] Step S401, initializing the network model parameters of the preset neural network model, or setting the network model parameters of the previous preset time length as the current network model parameters of the preset neural network model;

[0094] Step S402, obtaining a training data set including the text data, extracting the text data to obtain a plurality of word vector groups;

[0095] Step S403, inputting the word vector group corresponding to the preset time length into the preset neural network model, so that the preset neural network model adjusts the network model parameters to minimize the value of the loss function;

[0096] Step S404: Use word vector groups corresponding to multiple preset time lengths to train the preset neural network model, and determine the final network model parameters of the preset neural network model after the training is completed.

[0097] In the stock index compilation task, the neural network model can be trained using text data corresponding to multiple dates and stock price information.

[0098] First, the parameters of the neural network model can be initialized to D0, and the text data and stock price information corresponding to the first date of multiple dates are used to train the neural network model for the first time. The training goal is to minimize the value of the loss function L in formula (4):

[0099]

[0100] Among them, i1 and i2 are the serial numbers of any two companies. The "Newton method" and other methods can be used to adjust the network parameters by using the derivative of the loss function to the network parameters (c, v, h) to minimize the value of the loss function.

[0101] The meaning of the above loss function is that, with the stock price change rate as the supervisory information, for the same concept j, the weight of company i1 on concept j multiplied by the stock price change rate of company i1 on the same day should be as close as possible to the weight of company i2 on concept j multiplied by the stock price change rate of company i2 on the same day. That is, after the weights of different companies are adjusted and determined according to the knowledge graph, the impact on concept j should be consistent.

[0102] For example, according to statistics, there are three companies, Company A, Company B, and Company C, that involve the concept of games. The weight of the concept of games in the business of Company A is 70%, the weight of the concept of games in the business of Company B is 50%, and the weight of the concept of games in the business of Company C is 30%. Then the training goal of the neural network model is to adjust the network parameter D so that the stock price change rate of Company A*the weight of Company A, the stock price change rate of Company B*the weight of Company B, and the stock price change rate of Company C*the weight of Company C are as close as possible, and then determine the network parameters after this training.

[0103] The network parameters of the neural network model trained with the data of the first day can be named D1, and the network parameters of the neural network model trained with the data of the second day can be named D2, etc. The neural network parameters before training each day's data can copy the network parameters determined after training with the data of the previous day, as shown in formula (5):

[0104] D t =D t-1 …………………………………………(5)

[0105] Among them, t is the number of the date corresponding to the training data.

[0106] During the training process, you can use the Yard server to train model parameters and store sample data in the HDFS distributed storage environment.

[0107] After the training data set is fully trained, the model parameter D after training with the data of day T (T is the largest date number in the training data set) will be T Solidify into the parameters of the neural network model, that is, form a data processing model that can realize the data processing process of steps S1 to S4. It is worth mentioning that the network parameter D t At least including the word vector c i 、v a 、h j Network parameters for the three layers that perform processing.

[0108] In some embodiments, the knowledge graph structure may be complex, making the above neural network model and loss function very complex. For example, the business of Company A involves hundreds of concepts such as social networking, Internet advertising, mobile games, cloud, content, etc.

[0109] In one embodiment of the present disclosure, a random dropout structure may be used or a penalty term may be directly added to the loss function to simplify the network structure.

[0110] Dropout is a method for optimizing artificial neural networks with deep structures. During the learning process, some weights or outputs of the hidden layer are randomly reset to zero, reducing the co-dependence between nodes, thereby achieving regularization of the neural network and reducing its structural risk. There are many ways to implement random dropout according to the different structures of the neural network itself. For example, for multi-layer perceptrons (MLP), random dropout usually resets the output of the selected node to zero; for convolutional neural networks (CNN), random dropout can randomly reset some elements of the convolution kernel to zero, that is, random connection dropout, or randomly reset the channels of the entire feature map in the case of multiple channels, that is, spatial dropout; for recurrent neural networks (RNN), random dropout can act on the input and state matrix of each time step according to the topological structure of the network.

[0111] Adding a penalty term to the loss function, for example, multiplying the inverse of the square of all weights by an attenuation coefficient, and then calculating the sum of the attenuated weights, can penalize the smallest weight value so that the concept with a smaller weight value does not participate in the calculation. Of course, there are many ways to add penalty terms, and those skilled in the art can set the form of the penalty term according to the network structure and actual situation, and the present disclosure is not limited to this.

[0112] By simplifying the network structure by using random dropout and adding penalty items, the computational efficiency can be improved. For example, only several concepts with the largest weight (head) in company A's business are considered, such as Internet advertising and mobile games.

[0113] In another embodiment of the present disclosure, a neural network model can be used to perform a series of tasks such as text data capture, object dynamic index capture, word vector group extraction, weight calculation, category dynamic index extraction, etc. At this time, the neural network model is sequentially spliced ​​by the graph network model and the data processing model mentioned above, the input layer is used to receive text data and object dynamic indicators, and the output layer is used to output category dynamic indicators.

[0114] Since the neural network model needs to extract word vector groups, the loss function in the training process can be expressed as:

[0115]

[0116] Among them, Σ (i,a,j) l(c i ,v a ,h j ) is the loss function of the traditional knowledge graph mapping method (extracting word vectors from text data), that is, the loss function of the graph network model. It is the loss function of the dynamic index corresponding to the word vector calculation category in formula (4).

[0117] By using a data processing model to automatically capture and process text data, dynamic indicators of multiple categories can be automatically derived. The extensiveness of text data helps to identify new categories, and the analysis of text data by the data processing model helps to improve the accuracy of determining the weight of the object to the target category.

[0118] In summary, the embodiments of the present disclosure process a large amount of text data by using a graph network model, extracting the correspondence between objects and categories, and then calculating the weight of the object to the category. The dynamic indicators of multiple categories are determined based on the weight and the dynamic indicators of the object. This can timely identify new categories and improve the accuracy of judging the weight of the object to the category. It avoids the problems of untimely information and inaccurate labeling caused by the labeling of object categories in related technologies, and can effectively improve the accuracy of data mining.

[0119] Figure 5 It is a schematic diagram of an application scenario of an embodiment of the present disclosure.

[0120] refer to Figure 5 In the stock index compilation scenario, the trained neural network model can be used to perform the following processing to determine the real-time concept stock index change rate:

[0121] Step S51, capturing real-time text data containing the names of listed companies from sources such as WeChat public accounts and stock news information;

[0122] Step S52, obtaining real-time stock price information of multiple listed companies from the exchange platform, and calculating the stock price change rates of multiple listed companies based on the closing price information of the previous day and formula (1);

[0123] Step S53, using the neural network layer of the graph network model part, according to the "financial knowledge graph", the text data is processed into a triple word vector group in the form of (company i, behavior a, concept j);

[0124] Step S54, determining the weight of company i to concept j according to the triple word vector group and formula (3);

[0125] Step S55, determining the stock price change rate of concept j according to the above weight, stock price change rate and formula (2).

[0126] The above method can be used not only for index compilation of stock concepts, but also for index compilation of financial products such as stocks, futures, options, bonds, exchange rates, interest rate derivatives, etc. Of course, the method of the embodiment of the present disclosure is not limited to processing index compilation of financial products, but can also be applied to many other scenarios, and the present disclosure is not limited to this.

[0127] Corresponding to the above method embodiments, the present disclosure also provides a data processing device that can be used to execute the above method embodiments.

[0128] Figure 6 The block diagram schematically shows a data processing device in an exemplary embodiment of the present disclosure.

[0129] refer to Figure 6 , the data processing device 600 may include:

[0130] The text data acquisition module 602 is configured to acquire text data including objects.

[0131] The relational data extraction module 604 is configured to extract multiple word vector groups including the corresponding relationship between objects and categories from the text data.

[0132] The object dynamic index acquisition module 606 is configured to acquire the dynamic index of the object.

[0133] The category dynamic index determination module 608 is configured to determine the dynamic index of the category according to the multiple word vector groups and the dynamic index of the object.

[0134] In an exemplary embodiment of the present disclosure, the category dynamic index determination module 608 includes:

[0135] A weight determination unit 6081 is configured to determine weights of multiple objects for a target category based on the multiple word vector groups;

[0136] The comprehensive calculation unit 6082 is configured to determine the product of the weight of each object to the target category and the dynamic index of the object; take the sum of multiple products corresponding to multiple objects as the numerator; take the sum of the weights of the multiple objects to the target category as the denominator; and determine the dynamic index corresponding to the target category according to the ratio of the numerator to the denominator.

[0137] In an exemplary embodiment of the present disclosure, the weight determination unit 6081 is configured as:

[0138] Inputting the multiple word vector groups into a preset neural network model to obtain multiple output values;

[0139] Determine m output values ​​corresponding to the word vector group including both the target object and the target category from the multiple output values, where m≥1;

[0140] Determine n output values ​​corresponding to the word vector group including the target object from the multiple output values, where n≥1;

[0141] The weight of the target object to the target category is determined according to the ratio of the sum of the m output values ​​to the sum of the n output values.

[0142] In an exemplary embodiment of the present disclosure, the loss function of the preset neural network model includes:

[0143]

[0144] Among them, L is the loss function, j is the serial number of the target category, i1 and i2 are the serial numbers of the target objects, w is the weight of the target object to the target category, and x is the dynamic indicator of the target object.

[0145] In an exemplary embodiment of the present disclosure, a model training module 610 is further included, which is configured as follows:

[0146] Initializing the network model parameters of the preset neural network model, or setting the network model parameters of the previous preset time length as the current network model parameters of the preset neural network model;

[0147] Acquire a training data set including the text data, and extract the text data to obtain a plurality of word vector groups;

[0148] Inputting the word vector group corresponding to the preset time length into the preset neural network model, so that the preset neural network model adjusts the network model parameters to minimize the value of the loss function;

[0149] The preset neural network model is trained using word vector groups corresponding to multiple preset time lengths, and the final network model parameters of the preset neural network model are determined after the training is completed.

[0150] In an exemplary embodiment of the present disclosure, the object is a name of a listed company, the category is a stock concept, and the dynamic indicator is a stock price change rate.

[0151] In an exemplary embodiment of the present disclosure, the word vector group includes the corresponding relationship between the name of the listed company, the behavior and the stock concept.

[0152] Since the functions of the device 600 have been described in detail in the corresponding method embodiments, the present disclosure will not elaborate on them here.

[0153] It should be noted that, although several modules or units of the device for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. On the contrary, the features and functions of one module or unit described above can be further divided into multiple modules or units to be embodied.

[0154] In an exemplary embodiment of the present disclosure, an electronic device capable of implementing the above method is also provided.

[0155] It will be appreciated by those skilled in the art that various aspects of the present invention may be implemented as a system, method or program product. Therefore, various aspects of the present invention may be specifically implemented in the following forms, namely: a complete hardware implementation, a complete software implementation (including firmware, microcode, etc.), or a combination of hardware and software, which may be collectively referred to herein as a "circuit", "module" or "system".

[0156] Refer to the following Figure 7 An electronic device 700 according to this embodiment of the present invention will be described. Figure 7 The electronic device 700 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0157] like Figure 7As shown, the electronic device 700 is in the form of a general computing device. The components of the electronic device 700 may include but are not limited to: at least one processing unit 710, at least one storage unit 720, and a bus 730 connecting different system components (including the storage unit 720 and the processing unit 710).

[0158] The storage unit stores program codes, which can be executed by the processing unit 710, so that the processing unit 710 performs the steps according to various exemplary embodiments of the present invention described in the above “Exemplary Method” section of this specification. For example, the processing unit 710 can perform the following steps: Figure 1 The steps S1 shown in the figure are: obtaining text data including an object; step S2: extracting multiple word vector groups including the correspondence between objects and categories from the text data; step S3: obtaining the dynamic index of the object; step S4: determining the dynamic index of the category based on the multiple word vector groups and the dynamic index of the object.

[0159] The storage unit 720 may include a readable medium in the form of a volatile storage unit, such as a random access storage unit (RAM) 7201 and / or a cache storage unit 7202 , and may further include a read-only storage unit (ROM) 7203 .

[0160] The storage unit 720 may also include a program / utility 7204 having a set (at least one) of program modules 7205, such program modules 7205 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network model environment.

[0161] Bus 730 may represent one or more of several types of bus structures, including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.

[0162] The electronic device 700 may also communicate with one or more external devices 800 (e.g., keyboards, pointing devices, Bluetooth devices, etc.), may also communicate with one or more devices that enable a user to interact with the electronic device 700, and / or communicate with any device that enables the electronic device 700 to communicate with one or more other computing devices (e.g., routers, modems, etc.). Such communication may be performed via an input / output (I / O) interface 750. Furthermore, the electronic device 700 may also communicate with one or more network models (e.g., local area networks (LANs), wide area networks (WANs), and / or public network models, such as the Internet) via a network adapter 760. As shown, the network adapter 760 communicates with other modules of the electronic device 700 via a bus 730. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the electronic device 700, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, etc.

[0163] Through the description of the above implementation methods, it is easy for those skilled in the art to understand that the example implementation methods described here can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solution according to the implementation methods of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network model, including several instructions to enable a computing device (which can be a personal computer, a server, a terminal device, or a network model device, etc.) to execute the method according to the implementation methods of the present disclosure.

[0164] In an exemplary embodiment of the present disclosure, a computer-readable storage medium is also provided, on which a program product capable of implementing the above method of the present specification is stored. In some possible implementations, various aspects of the present invention can also be implemented in the form of a program product, which includes a program code, and when the program product is run on a terminal device, the program code is used to enable the terminal device to execute the steps according to various exemplary embodiments of the present invention described in the above "Exemplary Method" section of the present specification.

[0165] The program product for implementing the above method according to an embodiment of the present invention can adopt a portable compact disk read-only memory (CD-ROM) and include program code, and can be run on a terminal device, such as a personal computer. However, the program product of the present invention is not limited thereto, and in this document, a readable storage medium can be any tangible medium containing or storing a program, which can be used by or in combination with an instruction execution system, an apparatus or a device.

[0166] The program product may use any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0167] Computer readable signal media may include data signals propagated in baseband or as part of a carrier wave, in which readable program code is carried. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Readable signal media may also be any readable medium other than a readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0168] The program code embodied on the readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination of the foregoing.

[0169] Program code for performing the operations of the present invention may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user computing device through any type of network model, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., through the Internet using an Internet service provider).

[0170] In addition, the above-mentioned figures are only schematic illustrations of the processes included in the method according to an exemplary embodiment of the present invention, and are not intended to be limiting. It is easy to understand that the processes shown in the above-mentioned figures do not indicate or limit the time sequence of these processes. In addition, it is also easy to understand that these processes can be performed synchronously or asynchronously, for example, in multiple modules.

[0171] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art that are not disclosed in the present disclosure. The specification and examples are to be considered exemplary only, and the true scope and concept of the present disclosure are indicated by the claims.

Claims

1. A data processing method, characterized in that: include: Get text data including objects; Extracting a plurality of word vector groups including correspondences between objects and categories from the text data; Obtaining dynamic indicators of the object; Inputting the multiple word vector groups into a preset neural network model to obtain multiple output values; Determine m output values ​​corresponding to the word vector group including both the target object and the target category from the multiple output values, where m≥1; Determine n output values ​​corresponding to the word vector group including the target object from the multiple output values, where n≥1; Determining the weight of the target object to the target category according to the ratio of the sum of the m output values ​​to the sum of the n output values; Determining the product of the weight of each object for the target category and the dynamic index of the object; Taking the sum of multiple products corresponding to multiple objects as a numerator; Taking the sum of the weights of the multiple objects to the target category as the denominator; The dynamic index corresponding to the target category is determined according to the ratio of the numerator to the denominator.

2. The data processing method according to claim 1, characterized in that: The loss function of the preset neural network model includes: Among them, L is the loss function, j is the serial number of the target category, i1 and i2 are the serial numbers of the target objects, w is the weight of the target object to the target category, and x is the dynamic indicator of the target object.

3. The data processing method according to claim 2, characterized in that: The training process of the preset neural network model includes: Initializing the network model parameters of the preset neural network model, or setting the network model parameters of the previous preset time length as the current network model parameters of the preset neural network model; Acquire a training data set including the text data, and extract the text data to obtain a plurality of word vector groups; Inputting the word vector group corresponding to the preset time length into the preset neural network model, so that the preset neural network model adjusts the network model parameters to minimize the value of the loss function; The preset neural network model is trained using word vector groups corresponding to multiple preset time lengths, and the final network model parameters of the preset neural network model are determined after the training is completed.

4. A data processing device, characterized in that: include: A text data acquisition module, configured to acquire text data including an object; A relational data extraction module, configured to extract a plurality of word vector groups including corresponding relationships between objects and categories from the text data; An object dynamic index acquisition module, configured to acquire the dynamic index of the object; A category dynamic indicator determination module, configured to input the plurality of word vector groups into a preset neural network model to obtain a plurality of output values; Determine m output values ​​corresponding to the word vector group including both the target object and the target category from the multiple output values, where m≥1; Determine n output values ​​corresponding to the word vector group including the target object from the multiple output values, where n≥1; Determining the weight of the target object to the target category according to the ratio of the sum of the m output values ​​to the sum of the n output values; Determining the product of the weight of each object for the target category and the dynamic index of the object; Taking the sum of multiple products corresponding to multiple objects as a numerator; Taking the sum of the weights of the multiple objects to the target category as the denominator; The dynamic index corresponding to the target category is determined according to the ratio of the numerator to the denominator.

5. An electronic device, characterized in that: include: Memory; as well as A processor coupled to the memory, wherein the processor is configured to execute the data processing method according to any one of claims 1 to 3 based on instructions stored in the memory.

6. A computer-readable storage medium having a program stored thereon, wherein when the program is executed by a processor, the data processing method according to any one of claims 1 to 3 is implemented.

Citation Information

Patent Citations

  • Novel data mining method

    CN108427753A

  • Popularity information generation method and device

    CN108733706A

  • Credit strategy evaluation and selection method based on summer-prize ratio

    CN110288466A