Object prediction method, apparatus and electronic device

Abstract

The application discloses a kind of prediction method, device and electronic equipment of object.It relates to machine learning field or other related fields, the method comprises: obtaining the object information of to-be-predicted object, object information at least includes the fund flow direction chart data of to-be-predicted object and time series data, time series data is the transaction data corresponding to each time point of to-be-predicted object in preset length;Determine the fund flow information of to-be-predicted object based on object information, wherein the fund flow information represents the transaction information of to-be-predicted object through target agency transaction;The fund flow information of to-be-predicted object is processed based on target model, and prediction result is obtained, wherein the prediction result represents the influence degree of to-be-predicted object to target agency, and target model is generated according to the object information of multiple target objects.The application solves the technical problem of low prediction efficiency in predicting the influence degree of customer to agency in the prior art.

Description

Technical Field

[0001] This invention relates to the field of machine learning or other related fields, and more specifically, to a method, apparatus, and electronic device for predicting objects. Background Technology

[0002] At present, accepting public deposits is one of the basic businesses of commercial banks, and the interest income from deposits and loans is the main source of income for commercial banks. How to accurately develop and attract high-value customers is a key issue facing commercial banks' deposit business at present.

[0003] In current technologies, banks primarily assess a customer's future value through their financial assets. When building models for fund prediction, graph information such as fund flows and investment relationships among customers is often incorporated to reduce prediction errors. However, existing graph neural networks are not ideally suited for processing time-series information, and training them is extremely time-consuming. Time-series information is crucial for predicting a customer's future funds; ignoring it increases prediction errors. Therefore, current technologies using graph neural networks to predict customer funds suffer from low prediction efficiency.

[0004] There is currently no effective solution to the above problems. Summary of the Invention

[0005] This invention provides a method, apparatus, and electronic device for predicting objects, in order to at least solve the technical problem of low prediction efficiency in the prior art when predicting the degree of influence of customers on an organization.

[0006] According to one aspect of the present invention, a method for predicting an object is provided, comprising: acquiring object information of an object to be predicted, the object information including at least fund flow diagram data and time series data of the object to be predicted, the time series data being transaction data corresponding to each time point of the object to be predicted within a preset time period; determining fund flow information of the object to be predicted based on the object information, wherein the fund flow information represents transaction information of the object to be predicted conducting transactions through a target institution; and processing the fund flow information of the object to be predicted based on a target model to obtain a prediction result, wherein the prediction result represents the degree of influence of the object to be predicted on the target institution, and the target model is generated based on object information of multiple target objects.

[0007] Furthermore, the target model is generated by: acquiring object information of multiple target objects; training a preset model based on the object information to generate the target model.

[0008] Furthermore, after obtaining object information for multiple target objects, the method further includes: determining the number of features possessed by each target object from a first information set, wherein the first information set contains information on features possessed by objects among the multiple target objects whose influence on the target organization is less than a preset value; determining objects among the multiple target objects whose number of features is less than a preset number of features to obtain a first object set; deleting feature information corresponding to the first object set from the first information set to obtain a target information set; determining objects with missing features from the multiple target objects based on the target information set to obtain a second object set; obtaining feature information corresponding to the second object set from the target information set to obtain a second information set; adding features to the objects corresponding to the second information set to obtain an updated second information set, wherein the objects after adding features have at least a first target feature, the first target feature being the feature with the largest number of features among the features possessed by other target objects, and other target objects being objects among the multiple target objects excluding those with missing features; updating the target information set based on the updated second information set to obtain an updated target information set.

[0009] Furthermore, after obtaining object information for multiple target objects, the method further includes: identifying objects with missing data class features among the multiple target objects in the third information set, obtaining a third object set, wherein the third information set contains information on the features possessed by objects among the multiple target objects whose influence on the target organization is greater than a preset value; obtaining feature information corresponding to the third object set from the third information set, obtaining a fourth object information set; adding features to the objects corresponding to the fourth object information set, obtaining an updated fourth object information set, wherein the objects after adding features have at least a second target feature, and the data value corresponding to the second target feature is the average data of the features corresponding to the objects in the third information set; updating the third information set based on the updated fourth object information set, obtaining an updated third information set.

[0010] Furthermore, object information of multiple target objects is obtained, including merging the updated target information set and the updated third information set to obtain time series data.

[0011] Furthermore, after obtaining object information for multiple target objects, the method also includes: normalizing the time-series data to obtain multiple feature data; and performing feature selection on the multiple feature data to obtain multiple target feature data.

[0012] Furthermore, feature selection is performed on multiple feature data to obtain multiple target feature data, including: obtaining the number of features corresponding to multiple feature data; when the number of features is greater than a preset feature number threshold, calculating the similarity between multiple feature data and preset feature data; and determining multiple target feature data from multiple feature data based on the similarity.

[0013] Furthermore, acquiring object information for multiple target objects includes: sorting the objects corresponding to the updated third information set to obtain a sorting result; determining the first target object from the updated third information set based on the sorting result; determining objects related to the first target object from the updated target information set to obtain related objects; acquiring fund flow information between the first target object and related objects; determining the weight between the first target object and related objects based on the fund flow information; and determining fund flow diagram data based on the weight and the relationship.

[0014] According to another aspect of the present invention, an object prediction device is also provided, comprising: an acquisition module, configured to acquire object information of an object to be predicted, the object information including at least fund flow diagram data and time series data of the object to be predicted, the time series data being transaction data corresponding to each time point of the object to be predicted within a preset time period; a determination module, configured to determine fund flow information of the object to be predicted based on the object information, wherein the fund flow information represents transaction information of the object to be predicted conducting transactions through a target institution; and a processing module, configured to process the fund flow information of the object to be predicted based on a target model to obtain a prediction result, wherein the prediction result represents the degree of influence of the object to be predicted on the target institution, and the target model is generated based on object information of multiple target objects.

[0015] According to another aspect of the present invention, a computer-readable storage medium is also provided, wherein a computer program is stored in the computer-readable storage medium, wherein the computer program is configured to execute the above-described object prediction method at runtime.

[0016] According to another aspect of the present invention, an electronic device is also provided, the electronic device including one or more processors; a memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors are configured to execute the prediction method of the object described above during runtime.

[0017] According to another aspect of the present invention, a computer program product is also provided, including a computer program / instructions that, when executed by a processor, implement the above-described method for predicting the object.

[0018] In this embodiment of the invention, a method is adopted to obtain the prediction result by processing the fund flow information of the object to be predicted based on the target model. First, the object information of the object to be predicted is obtained, and the fund flow information of the object to be predicted is determined based on the object information. Then, the fund flow information of the object to be predicted is processed based on the target model to obtain the prediction result. The fund flow information represents the transaction information of the object to be predicted through the target institution, and the prediction result represents the degree of influence of the object to be predicted on the target institution. The target model is generated based on the object information of multiple target objects. The object information includes at least the fund flow diagram data and time series data of multiple target objects. The time series data is the transaction data corresponding to each time point of the target object within a preset time period.

[0019] In the above process, the target model is generated by training based on the fund flow graph data and time series data of multiple target objects. Since the existing models are not well suited to processing time series information, resulting in large errors in the prediction results, training the target model with time series information can make the target model more stable and reduce the error of the model's prediction results. At the same time, by constructing fund flow graph data and incorporating graph data as much as possible as possible as input features for model training, the error of the prediction results is reduced, thereby improving the prediction efficiency of customer funds.

[0020] Therefore, the solution provided in this application achieves the purpose of predicting the capital flow information of the object to be predicted based on the target model, thereby improving the technical effect of predicting the object to be predicted, and thus solving the technical problem of low prediction efficiency in the existing technology for predicting the degree of influence of customers on institutions. Attached Figure Description

[0021] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0022] Figure 1 This is a schematic diagram of an optional object prediction method according to an embodiment of the present invention;

[0023] Figure 2 This is a schematic diagram of an optional object prediction method according to an embodiment of the present invention;

[0024] Figure 3 This is a schematic diagram of an optional process for mapping the flow of funds among multiple target objects according to an embodiment of the present invention;

[0025] Figure 4 A schematic diagram of an optional object prediction device according to an embodiment of the present invention;

[0026] Figure 5 A schematic diagram of an optional electronic device according to an embodiment of the present invention. Detailed Implementation

[0027] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0028] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0029] It should be noted that all relevant information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for display, data used for analysis, etc.) involved in this invention are information and data authorized by the user or fully authorized by all parties. For example, this system has an interface with the relevant user or organization. Before obtaining relevant information, it needs to send an acquisition request to the aforementioned user or organization through the interface, and obtain the relevant information after receiving consent from the aforementioned user or organization.

[0030] Example 1

[0031] According to an embodiment of the present invention, an embodiment of a method for predicting an object is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0032] Figure 1 This is a schematic diagram of a method for predicting objects according to an embodiment of the present invention, such as... Figure 1 As shown, the method includes the following steps:

[0033] Step S101: Obtain object information of the object to be predicted. The object information includes at least the capital flow chart data and time series data of the object to be predicted. The time series data is the transaction data of the object to be predicted at each time point within a preset time period.

[0034] In step S101, object information of the object to be predicted can be obtained based on devices such as systems, servers, and electronic devices. In this embodiment, the object information of the object to be predicted can be obtained through the system. The object to be predicted can be a customer, such as a bank's customer.

[0035] Optionally, the system obtains object information of the object to be predicted through a preset database. The object information includes data such as basic customer information, financial information, and investment information. Based on the above data, the system constructs a capital flow diagram and obtains the corresponding time series data.

[0036] Step S102: Determine the fund flow information of the object to be predicted based on the object information, wherein the fund flow information represents the transaction information of the object to be predicted through the target institution.

[0037] In step S102, the fund flow information includes, but is not limited to, transaction information such as fund information and investment information of the object to be predicted. Different target institutions (e.g., banks) correspond to different preset databases.

[0038] Step S103: Process the fund flow information of the object to be predicted based on the target model to obtain the prediction result. The prediction result represents the degree of influence of the object to be predicted on the target institution. The target model is generated based on the object information of multiple target objects.

[0039] In step S102, the system processes the fund flow information of the object to be predicted using the trained target model to obtain the prediction result. In this embodiment, after obtaining the prediction result, the target model can compare the output prediction result with a preset fund threshold. If the prediction result is greater than the preset fund threshold, the system outputs that the object to be predicted is an object with a high degree of influence on the target institution. For example, if the prediction result is greater than the preset fund threshold, the system determines that the object to be predicted is a high-value customer of the bank, and then stores the basic information of the high-value customer in a preset high-value customer database.

[0040] Optionally, the target model is generated by training based on object information from multiple target objects, and is an LSTM (Long-Short Term Memory) model. The model is generated as follows: Figure 2As shown, by preprocessing the object information of multiple target objects obtained from a preset database, fund flow diagram data and time series data are obtained. Then, after standardizing the data, feature selection is performed on the data features. When the number of features reaches a preset feature number threshold, target feature data is obtained. Finally, an LSTM network model is constructed, and the parameters of the LSTM network model are optimized. Training is started based on the fund flow diagram data and target feature data to obtain the target LSTM network model. Based on the target LSTM network model, the prediction of the object to be predicted is performed, and the prediction result is output. When the object to be predicted is a high-value object, the prediction result is saved to a preset high-value customer database.

[0041] Based on the scheme defined in steps S101 to S103 above, it can be understood that in this embodiment of the invention, the method of processing the fund flow information of the object to be predicted based on the target model to obtain the prediction result firstly obtains the object information of the object to be predicted, determines the fund flow information of the object to be predicted based on the object information, and then processes the fund flow information of the object to be predicted based on the target model to obtain the prediction result; wherein, the fund flow information represents the transaction information of the object to be predicted through the target institution, the prediction result represents the degree of influence of the object to be predicted on the target institution, the target model is generated based on the object information of multiple target objects, the object information includes at least the fund flow diagram data and time series data of multiple target objects, and the time series data is the transaction data corresponding to each time point of the target object within a preset time period.

[0042] It is noteworthy that in the above process, the target model is trained and generated based on the fund flow graph data and time series data of multiple target objects. Since the existing models are not well suited to processing time series information, resulting in large errors in the prediction results, training the target model with time series information can make the target model more stable and reduce the error of the model's prediction results. At the same time, by constructing fund flow graph data and incorporating graph data as much as possible as possible as the input features for model training, the error of the prediction results is reduced, thereby improving the prediction efficiency of customer funds.

[0043] Therefore, the solution provided in this application achieves the purpose of predicting the capital flow information of the object to be predicted based on the target model, thereby improving the technical effect of predicting the object to be predicted, and thus solving the technical problem of low prediction efficiency in the existing technology for predicting the degree of influence of customers on institutions.

[0044] In one optional embodiment, the target model in the system is generated by: acquiring object information of multiple target objects; training a preset model based on the object information to generate the target model.

[0045] In this embodiment, the system first obtains the object information of the target object, such as basic information, financial information, investment information and other data information of the target object. Then, the object information is preprocessed and standardized. After that, feature selection is performed on the data features to obtain input feature data. The input feature data can be divided into training set and test set for training. For example, the training set and test set are formed according to the date feature. 12 months of customer information data are selected, with the first 9 months as the training set train_data and the last 3 months as the test set test_data.

[0046] Secondly, construct the LSTM network model. LSTM is a special type of recurrent neural network model, which incorporates a gating mechanism and a new internal state c. t The internal state at time t can solve the gradient explosion or vanishing problem in traditional recurrent neural networks. Its gating mechanism includes a forget gate, an input gate, and an output gate, which can be described by the formula:

[0047]

[0048]

[0049] c t =f1*c t-1 +f2(3)

[0050] Where f1 is the mathematical expression for the forget gate, which ignores unimportant information; f2 is the mathematical expression for the input gate, which selects important information for input; and χ... t For event information, c t c represents the internal state at time t. t In addition to being passed down, it will also be used to update the current short-term memory. t Multiple c t It can form a long-term memory chain, and s t The short-term memory chains update each other, and finally output the predicted value y. t In the above formula, w and b represent the weights and biases of the network in each iteration, respectively, and sigmoid is a commonly used activation function in neural networks with a range of [0,1].

[0051] Furthermore, the parameters of the LSTM network model are fine-tuned. The system adjusts the learning rate, number of network layers, number of neurons per layer, and batch size of the LSTM network model to optimize it. For example, the adaptive learning rate adjustment algorithm AdaDelta can be used to adjust the learning rate, which can smooth out fluctuations in the learning rate and improve the learning speed and accuracy to a certain extent. The number of network layers, number of neurons per layer, and batch size are adjusted through grid search.

[0052] Furthermore, after fine-tuning the model's parameters, the system inputs the feature data from the training set into the LSTM network model for training, and evaluates the model using metrics such as Symmetric Mean Percentage Error (SMAPE). The formula for calculating SMAPE is as follows:

[0053]

[0054] in Represents the predicted value, y i This represents the true value, and n represents the number of samples.

[0055] It should be noted that SMAPE is used as an evaluation metric because it can correct for MAPE and avoid the error rate from becoming too large due to the true value being too small.

[0056] Finally, the system inputs the test samples into the LSTM network model to test it, obtaining the trained target model. For example, inputting the test sample `test_data` into the LSTM network model yields the predicted financial value for the corresponding customer over the next three months, and then measures the customer's value according to a value standard. Optionally, after obtaining the target model, it is stored in a pre-defined model database.

[0057] It should be noted that training the preset model with the object information of the target object improves the model's learning performance and effectiveness, thereby enhancing the target model's prediction efficiency for the object's funds.

[0058] Furthermore, after acquiring object information for multiple target objects, the method further includes: firstly, determining the number of features possessed by each target object from a first information set, wherein the first information set contains information on features possessed by objects whose influence on the target organization is less than a preset value; then, determining objects with fewer than a preset number of features from the multiple target objects to obtain a first object set; and deleting the feature information corresponding to the first object set from the first information set to obtain a target information set; secondly, determining objects with missing features from the multiple target objects based on the target information set to obtain a second object set; then, acquiring the feature information corresponding to the second object set from the target information set to obtain a second information set; finally, adding features to the objects corresponding to the second information set to obtain an updated second information set, wherein the objects with added features possess at least a first target feature, which is the feature with the most features among the features possessed by other target objects, and other target objects are objects among the multiple target objects excluding those with missing features; and updating the target information set based on the updated second information set to obtain an updated target information set.

[0059] In this embodiment, the first information set can be the object information of the target object whose influence on the target institution is less than a preset value, such as the low-value customer information set of a bank. By filtering and filling the data in the first information set, the data of objects with too many missing features and objects lacking credit scores in the first information set are deleted. For objects lacking fewer feature data, the feature with the most features among the features of other target objects is added.

[0060] Optionally, firstly, the first object set consists of objects in the first information set that lack most of their feature data. Invalid information in the first information set can be deleted based on the first object set to obtain the target information set. Secondly, some objects in the target information set still lack feature data. By detecting objects in the target information set that lack features, a second object set is determined. The features corresponding to the objects in the second object set are obtained from the target information set to determine the second information set. Features are added based on the feature with the most features among the features possessed by other target objects to obtain the updated second information set. The target information set is then updated based on the updated second information set to obtain the updated target information set.

[0061] It should be noted that by filtering and filling the information in the first information set, the effectiveness of the model training set data is improved, thereby improving the accuracy of the target model's prediction results and thus improving the target model's prediction efficiency for the target's funds.

[0062] Furthermore, after acquiring object information for multiple target objects, the system first identifies objects with missing data-type features from the third information set, thus obtaining a third object set. This third information set contains information about the features possessed by objects whose influence on the target organization exceeds a preset value. Then, it extracts the feature information corresponding to the third object set from the third information set, resulting in a fourth object information set. Next, features are added to the objects in the fourth object information set, resulting in an updated fourth object information set. The objects with added features possess at least a second target feature, and the data value corresponding to the second target feature is the average data of the corresponding features in the third information set. Based on the updated fourth object information set, the third information set is updated, resulting in an updated third information set.

[0063] In this embodiment, the third information set is a set of information about the characteristics of objects whose influence on the target organization is greater than a preset value, such as a set of high-value customer information of a bank. By filling the third information set with the average data of the corresponding characteristics of the objects in the third information set, objects with missing data type characteristics are filled with feature data, and the data-processed third information set is obtained, that is, the updated third information set.

[0064] Optionally, firstly, the third object set is the set corresponding to objects in the third information set that lack some data class features. Based on the third information set, the feature information corresponding to the objects in the third object set can be determined to obtain the fourth object information set. Then, based on the average data of the features corresponding to the objects in the third information set, the missing features in the fourth object information set are filled to obtain the updated fourth object information set. Then, based on the updated fourth object information set, the third object set is updated to obtain the third information set after missing data filling, i.e., the updated third information set.

[0065] It should be noted that by filling in the missing data in the third information set, the effectiveness of the model training set data, the accuracy of the target model's prediction results, and the target model's prediction efficiency for the target's funds are further improved.

[0066] In another optional embodiment, the system obtains the number of days between two dates by setting a date base value and calculating the difference between the dataset date and the date base value. The construction method is as follows:

[0067] base_time=min(timestap(data_time))

[0068] sub_time=time(timestap(data_time)-base_time)

[0069] Here, `data_time` is the date column in the dataset. This value is converted to a timestamp, and the smallest timestamp, `base_time`, is taken as the base date value. Then, the difference between the timestamps in the dataset and the base date timestamp is calculated to obtain the number of days between the two dates, `sub_time`. The timestamp is used to authenticate the time when the data was generated. By calculating the difference between the base date value and the dataset dates, time calculations are facilitated, as well as the visualization and time-series representation of subsequent prediction results.

[0070] Furthermore, the system acquires object information from multiple target objects, including merging the updated target information set and the updated third information set to obtain time-series data.

[0071] In this embodiment, the system merges the high-value dataset and low-value dataset after filtering and filling to obtain a time-series dataset for subsequent training of the target model.

[0072] Furthermore, after acquiring object information for multiple target objects, the system normalizes the time-series data to obtain multiple feature data; then, it performs feature selection on the multiple feature data to obtain multiple target feature data.

[0073] In this embodiment, the system obtains feature data by normalizing the time series data, and then selects features from the feature data to obtain target feature data, wherein the target feature data is used to train the target model.

[0074] Optionally, normalization can be performed on the data in the time series data excluding the funds column, adjusting each feature dimension to have a mean of 0 and a variance of 1. The calculation method for normalization is as follows:

[0075]

[0076] Where x is the data value, σ is the standard deviation, and μ is the mean.

[0077] It should be noted that by normalizing the data, the convergence of the data is accelerated, the training time of the model is reduced, and the prediction efficiency is further improved.

[0078] Furthermore, the system obtains the number of features corresponding to multiple feature data; then, when the number of features exceeds a preset feature number threshold, it calculates the similarity between the multiple feature data and the preset feature data; and determines multiple target feature data from the multiple feature data based on the similarity.

[0079] In this embodiment, the chi-square hypothesis is used for feature selection. Feature selection is performed by calculating the chi-square score of each feature. The formula for calculating the chi-square hypothesis is as follows:

[0080]

[0081] Where A is the actual value and E is the expected value.

[0082] Optionally, the higher the chi-square value, the higher the correlation between the actual value and the expected value of the feature. In this embodiment, features with larger chi-square values ​​are selected, and the calculation is stopped when the number of features with larger chi-square values ​​reaches a preset feature number threshold, and the target feature data is output.

[0083] It should be noted that feature selection of data can remove irrelevant features, thereby simplifying feature data and reducing model training time.

[0084] Alternatively, when acquiring object information for multiple target objects, the system sorts the objects corresponding to the updated third information set to obtain a sorting result; based on the sorting result, it determines the first target object from the updated third information set; it determines the objects that are related to the first target object from the updated target information set to obtain related objects; it acquires the fund flow information between the first target object and the related objects; it determines the weight between the first target object and the related objects based on the fund flow information; and it determines the fund flow diagram data based on the weight and the relationship.

[0085] In this embodiment, objects in the high-value customer set corresponding to the third information set are sorted according to value criteria, such as sorting from high to low value or from low to high value; then an object with high value is determined, for example, the object ranked first in value sorting is determined as the first target object.

[0086] Furthermore, such as Figure 3As shown, the length of the time-series data `data_len` and the length of the high-value customer dataset `data2_len` are obtained, and adjacency matrices `m1` and `m2` of size `[data_len, data2_len]` are created. Then, the primary key of `data2` is traversed, and the primary key of the high-value customer dataset is queried in the time-series data using a preset SQL script to obtain customers with financial transactions with the current high-value customer. At this point, using this high-value customer as the center point, other customers with financial transactions with it are continuously connected, with funds as the weight. The networkx database is used to convert this into a fund flow graph. The fund flow graph is then continuously updated in a loop until the last high-value customer node is traversed, and the adjacency matrix of the finally updated fund flow graph is transferred to matrix `m1`, and the node information is saved to matrix `m2`. The Floya algorithm is used to continuously update the weights of matrix `m1`. To prevent the weights between two indirectly connected points from increasing, the weights are the sum of funds between the two indirectly connected points / the number of nodes, until the algorithm ends.

[0087] It should be noted that by constructing a fund flow graph, the graph data can be used as input features for model training, which can enable the model to output more reliable prediction results with less empirical risk.

[0088] Therefore, this invention provides a new method for predicting objects. This method, through data preprocessing, feature selection, model construction and training, obtains a model with higher efficiency and more stable results with smaller errors. Compared with graph neural networks, it is more suitable for processing time-series data, and it pays more attention to the relationship between customers and financial transactions than the traditional LSTM model. This solves the technical problem of low prediction efficiency in predicting the degree of influence of customers on institutions in the existing technology.

[0089] Example 2

[0090] According to an embodiment of the present invention, an embodiment of an object prediction device is provided, wherein, Figure 4 A schematic diagram of an optional object prediction device according to an embodiment of the present invention, such as... Figure 4 As shown, the device includes:

[0091] The acquisition module 401 is used to acquire object information of the object to be predicted. The object information includes at least the capital flow diagram data and time series data of the object to be predicted. The time series data is the transaction data corresponding to each time point of the object to be predicted within a preset time period. The determination module 402 is used to determine the capital flow information of the object to be predicted based on the object information. The capital flow information represents the transaction information of the object to be predicted through the target institution. The processing module 403 is used to process the capital flow information of the object to be predicted based on the target model to obtain the prediction result. The prediction result represents the degree of influence of the object to be predicted on the target institution. The target model is generated based on the object information of multiple target objects.

[0092] It should be noted that the above-mentioned acquisition module 401, determination module 402 and processing module 403 correspond to steps S101 to S103 in the above embodiments. The examples and application scenarios implemented by the three modules and the corresponding steps are the same, but are not limited to the content disclosed in the above embodiment 1.

[0093] Optionally, the object prediction device further includes: a first acquisition module and a first generation module; the first acquisition module is used to acquire object information of multiple target objects; the first generation module is used to train a preset model based on the object information to generate a target model.

[0094] Optionally, the object prediction device further includes: a first determining module, a second determining module, a deletion module, a third determining module, a second acquiring module, a first adding module, and a first updating module; the first determining module is used to determine the number of features of each target object from a first information set, wherein the first information set contains information on the features of objects whose influence on the target mechanism is less than a preset value among multiple target objects; the second determining module is used to determine objects whose number of features is less than a preset number of features from multiple target objects, thereby obtaining a first object set; the deletion module is used to delete the feature information corresponding to the first object set from the first information set, thereby obtaining a target information set; the third determining module is used to determine the number of features of each target object from the target information set. The first information set identifies objects with missing features from multiple target objects, resulting in a second object set. The second acquisition module acquires feature information corresponding to the second object set from the target information set, resulting in a second information set. The first addition module adds features to the objects corresponding to the second information set, resulting in an updated second information set. The objects with added features have at least a first target feature, which is the feature with the most features among the features possessed by other target objects. The other target objects are objects other than those with missing features. The first update module updates the target information set based on the updated second information set, resulting in an updated target information set.

[0095] Optionally, the object prediction device further includes: a fourth determining module, a third acquiring module, a second adding module, and a second updating module; the fourth determining module is used to determine objects with missing data class features among multiple target objects in the third information set, to obtain a third object set, wherein the third information set contains information on the features possessed by objects among multiple target objects whose influence on the target mechanism is greater than a preset value; the third acquiring module is used to acquire feature information corresponding to the third object set from the third information set, to obtain a fourth object information set; the second adding module is used to add features to the objects corresponding to the fourth object information set, to obtain an updated fourth object information set, wherein the objects after adding features have at least a second target feature, and the data value corresponding to the second target feature is the average data of the features corresponding to the objects in the third information set; the second updating module is used to update the third information set based on the updated fourth object information set, to obtain an updated third information set.

[0096] Optionally, the object prediction device further includes: a merging module; the merging module is used to merge the updated target information set and the updated third information set to obtain time series data.

[0097] Optionally, the object prediction device further includes a processing module and a selection module. The processing module is used to normalize the time-series data to obtain multiple feature data; the selection module is used to select features from the multiple feature data to obtain multiple target feature data.

[0098] Optionally, the object prediction device further includes: a fourth acquisition module, a calculation module, and a fifth determination module; the fourth acquisition module is used to acquire the number of features corresponding to multiple feature data; the calculation module is used to calculate the similarity between multiple feature data and preset feature data when the number of features is greater than a preset feature number threshold; the fifth determination module is used to determine multiple target feature data from multiple feature data based on the similarity.

[0099] Optionally, the object prediction device further includes: a sorting module, a sixth determining module, a seventh determining module, a fifth acquiring module, an eighth determining module, and a ninth determining module; the sorting module is used to sort the objects corresponding to the updated third information set to obtain a sorting result; the sixth determining module is used to determine the first target object from the updated third information set based on the sorting result; the seventh determining module is used to determine the objects that have a relationship with the first target object from the updated target information set to obtain the associated objects; the fifth acquiring module is used to acquire the fund flow information between the first target object and the associated objects; the eighth determining module is used to determine the weight between the first target object and the associated objects based on the fund flow information; and the ninth determining module is used to determine the fund flow diagram data based on the weights and the relationship.

[0100] Example 3

[0101] According to another aspect of the present invention, a computer-readable storage medium is also provided, wherein a computer program is stored in the computer-readable storage medium, wherein the computer program is configured to execute the above-described object prediction method at runtime.

[0102] Example 4

[0103] According to another aspect of the present invention, an electronic device is also provided, wherein, Figure 5 This is a schematic diagram of an optional electronic device according to an embodiment of the present invention, such as... Figure 5 As shown, the electronic device includes one or more processors; and a memory for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement a method for running the program, wherein the program is configured to execute the predicted method of the object described above during runtime.

[0104] Example 5

[0105] According to another aspect of the present invention, a computer program product is also provided, including a computer program / instructions that, when executed by a processor, implement the above-described method for predicting the object.

[0106] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0107] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0108] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some interfaces; indirect couplings or communication connections between units or modules may be electrical or other forms.

[0109] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0110] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0111] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, 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. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0112] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for predicting an object, characterized in that, include: Obtain object information of the object to be predicted. The object information includes at least the capital flow chart data and time series data of the object to be predicted. The time series data is the transaction data of the object to be predicted at each time point within a preset time period. Based on the object information, the fund flow information of the object to be predicted is determined, wherein the fund flow information represents the transaction information of the object to be predicted through the target institution; The fund flow information of the object to be predicted is processed based on the target model to obtain the prediction result, wherein the prediction result characterizes the degree of influence of the object to be predicted on the target institution, and the target model is generated based on the object information of multiple target objects; The fund flow chart data is constructed in the following manner: The objects corresponding to the updated third information set are sorted to obtain the sorting result. The third information set is a set of information on the characteristics of objects whose influence on the target organization is greater than a preset value. The updated third information set is obtained by filling the feature data of objects that are missing data type characteristics in the third information set based on the average data of the corresponding characteristics of the objects in the third information set. The first target object is determined from the updated third information set based on the sorting result; From the updated target information set, determine the objects that are associated with the first target object to obtain the associated objects; Obtain the fund transfer information between the first target object and the associated object; The weight between the first target object and the associated object is determined based on the fund flow information; The fund flow diagram data is determined based on the weights and the relationships.

2. The method according to claim 1, characterized in that, The target model is generated in the following way: Obtain object information of the multiple target objects; The preset model is trained based on the object information to generate the target model.

3. The method according to claim 2, characterized in that, After obtaining the object information of the plurality of target objects, the method further includes: The number of features possessed by each target object is determined from the first information set, wherein the first information set includes information on the features possessed by the objects among the plurality of target objects whose influence on the target organization is less than a preset value; From the plurality of target objects, objects with a number of features less than a preset number of features are determined to obtain a first object set; Remove the feature information corresponding to the first object set from the first information set to obtain the target information set; Based on the target information set, objects with missing features are determined from the plurality of target objects to obtain a second object set; The feature information corresponding to the second object set is obtained from the target information set to obtain the second information set; Features are added to the objects corresponding to the second information set to obtain an updated second information set. The objects after adding features have at least a first target feature. The first target feature is the feature with the most features among the features possessed by other target objects. The other target objects are the objects among the plurality of target objects excluding those with missing features. Based on the updated second information set, the target information set is updated to obtain the updated target information set.

4. The method according to claim 2, characterized in that, After obtaining the object information of the plurality of target objects, the method further includes: In the third information set, objects with missing data class characteristics among the multiple target objects are identified to obtain a third object set, wherein the third information set contains information on the characteristics of objects among the multiple target objects whose influence on the target organization is greater than a preset value; The feature information corresponding to the third object set is obtained from the third information set to obtain the fourth object information set; Features are added to the objects corresponding to the fourth object information set to obtain an updated fourth object information set, wherein the objects after adding features have at least a second target feature, and the data value corresponding to the second target feature is the average data of the features corresponding to the objects in the third information set; Based on the updated fourth object information set, the third information set is updated to obtain the updated third information set.

5. The method according to claim 4, characterized in that, Obtaining object information of the multiple target objects includes: The updated target information set and the updated third information set are merged to obtain the time series data.

6. The method according to claim 5, characterized in that, After obtaining the object information of the plurality of target objects, the method further includes: The time-series data is normalized to obtain multiple feature data. Feature selection is performed on the multiple feature data to obtain multiple target feature data.

7. The method according to claim 6, characterized in that, Feature selection is performed on the multiple feature data to obtain multiple target feature data, including: Obtain the number of features corresponding to the multiple feature data; When the number of features is greater than a preset feature number threshold, the similarity between the multiple feature data and the preset feature data is calculated; The multiple target feature data are determined from the multiple feature data based on the similarity.

8. An object prediction device, characterized in that, include: The acquisition module is used to acquire object information of the object to be predicted. The object information includes at least the capital flow diagram data and time series data of the object to be predicted. The time series data is the transaction data of the object to be predicted at each time point within a preset time period. The determination module is used to determine the fund flow information of the object to be predicted based on the object information, wherein the fund flow information represents the transaction information of the object to be predicted through the target institution; The processing module is used to process the capital flow information of the object to be predicted based on the target model to obtain a prediction result, wherein the prediction result characterizes the degree of influence of the object to be predicted on the target institution, and the target model is generated based on the object information of multiple target objects; The device is further configured to: sort the objects corresponding to the updated third information set to obtain a sorting result; wherein the third information set is a set of information on the characteristics of objects whose influence on the target institution is greater than a preset value; the updated third information set is obtained by filling in the missing data type characteristics of objects in the third information set with the average data of the corresponding characteristics of the objects in the third information set; determine a first target object from the updated third information set based on the sorting result; determine objects that are associated with the first target object from the updated target information set to obtain associated objects; obtain fund flow information between the first target object and the associated objects; determine the weight between the first target object and the associated objects based on the fund flow information; and determine the fund flow diagram data based on the weight and the association relationship.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, wherein the computer program is configured to execute the prediction method for the object as described in any one of claims 1 to 7 when it is run.

10. An electronic device, characterized in that, The electronic device includes one or more processors; A memory for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to be configured to run the programs, wherein the programs are configured to execute the prediction method for the object as described in any one of claims 1 to 7 at runtime.

11. A computer program product, comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the prediction method for the object as described in any one of claims 1 to 7.

Images