Financial product screening method and device, electronic equipment and computer program product

Abstract

The invention discloses a financial product screening method and device, electronic equipment and a computer program product. The method comprises the steps of obtaining user information of a financial user, determining product browsing data of the financial user in a historical time period based on the user information, and obtaining historical product browsing data; the historical product browsing data is input into a product screening model, a financial product recommendation list is output, the product screening model is obtained by training a graph convolutional neural network model through a training set, the graph convolutional neural network model is obtained by constructing a target graph model, and the target graph model is obtained by a user; the financial product recommendation list comprises a plurality of candidate financial products and prediction scores associated with the candidate financial products; and screening a preset number of candidate financial products from the financial product recommendation list to obtain a target financial product, and sending the target financial product to a client used by the financial user. Through the method and the device, the technical problem of low screening accuracy during financial product screening in related technologies is solved.

Description

Technical Field

[0001] This application relates to the field of financial technology, and more specifically, to a method, apparatus, electronic device, and computer program product for screening financial products. Background Technology

[0002] In the current era of digital finance, financial institutions are facing an ever-growing demand for personalized financial services. Traditional financial product recommendation methods are often too generic and cannot meet market demands or accurately match the interests and risk tolerance of financial users. To address this, personalized recommendation algorithms based on user behavior analysis have emerged. Among them, the most widely used is the collaborative filtering algorithm. The collaborative filtering algorithm uses users' historical behavior data to mine the similarity between potential user interests and products, thereby achieving personalized services.

[0003] In related technologies, collaborative filtering algorithms can predict a user's preference for products they haven't encountered before by analyzing user behavior. They can be divided into two main categories: user-based collaborative filtering and item-based collaborative filtering. User-based collaborative filtering requires collecting each user's ratings of various financial products, then calculating the similarity between users based on their shared ratings, and finally recommending financial products based on positive reviews from similar users. Item-based collaborative filtering, on the other hand, calculates the similarity between different financial products based on user ratings and recommends similar financial products to the current user.

[0004] However, collaborative filtering-based financial product recommendation algorithms have significant drawbacks. For example, for newly added users or newly launched financial products, the lack of sufficient historical behavioral data makes it difficult to establish effective user-product associations, resulting in poor performance. As the user base and the number of financial products grow, the user-product interaction matrix becomes extremely sparse, affecting the accuracy and diversity of recommendations. Collaborative filtering tends to favor popular financial products, which can easily lead to homogenization and uniformity in recommendations, limiting the breadth and depth of user experience.

[0005] There is currently no effective solution to the technical problem of low accuracy in screening financial products using related technologies. Summary of the Invention

[0006] The main objective of this application is to provide a method, apparatus, electronic device, and computer program product for screening financial products, so as to solve the technical problem of low screening accuracy in the related art when screening financial products.

[0007] To achieve the above objectives, according to one aspect of this application, a method for screening financial products is provided. The method includes: obtaining user information of a financial user; determining product browsing data of the financial user over a historical time period based on the user information to obtain historical product browsing data; inputting the historical product browsing data into a product screening model to output a list of recommended financial products, wherein the product screening model is trained on a graph convolutional neural network model using a training set, the graph convolutional neural network model is constructed from a target graph model, the training set consists of historical product browsing records of multiple financial users and the screened financial products associated with each financial user, and the recommended financial product list includes multiple candidate financial products and a predicted score associated with each candidate financial product; selecting a preset number of candidate financial products from the recommended financial product list to obtain target financial products, and sending the target financial products to the client used by the financial user.

[0008] Furthermore, the product selection model is trained as follows: A list of financial products is obtained, along with browsing records associated with M financial products in the list, resulting in M ​​sets of browsing records, where M is a positive integer; for each financial product, Y related financial products are selected from the set of browsing records associated with that financial product, resulting in M ​​sets of related financial products, where Y is a positive integer; a graph model is constructed based on the M sets of related financial products and the M financial products to obtain the target graph model, and an initial product selection model is built based on the target graph model; historical product browsing records of K financial users and the selected financial products associated with each financial user are obtained, and a training set is constructed from the K historical product browsing records and the K selected financial products, where K is a positive integer; the initial product selection model is trained using the training set to obtain the final product selection model.

[0009] Furthermore, the browsing records associated with M financial products in the list of purchased financial products are obtained, resulting in M ​​sets of browsing records. These include: obtaining the financial users who purchased each financial product in the list of purchased financial products, resulting in M ​​sets of financial users, where each set of financial users includes N financial users, and N is a positive integer; for each financial product, the browsing records of a set of financial users corresponding to each financial product before purchasing each financial product are obtained, resulting in M ​​sets of browsing records, where each set of browsing records includes R browsing records, and R is a positive integer.

[0010] Furthermore, selecting Y financial products associated with each financial product from a set of browsing records associated with that financial product includes: for a financial product, selecting Q browsed financial products from each browsing record associated with the financial product to obtain R sets of browsed financial products, where Q is a positive integer; determining the frequency data of each browsed financial product in the R sets of browsed financial products, and filtering the R sets of browsed financial products based on the frequency data of each browsed financial product to obtain Y browsed financial products with the highest frequency data; and identifying the Y browsed financial products with the highest frequency data as the Y financial products.

[0011] Furthermore, based on the M groups of related financial products and the M financial products themselves, a graph model is constructed to obtain the target graph model, which includes: for the M financial products, obtaining the product feature information of each financial product to obtain M*Y product feature information, wherein each product feature information includes at least one of the following: product type, risk level, screening popularity, and release time; the financial products are identified as nodes, the group of related financial products corresponding to the financial products are identified as adjacent nodes of the nodes, and the relationship between the financial products and each associated financial product is identified as an edge; based on the M nodes, the M groups of adjacent nodes, the M groups of edges, and the M*Y product feature information, a graph model is constructed to obtain the target graph model.

[0012] Furthermore, a graph model is constructed based on M nodes, M groups of adjacent nodes, M groups of edges, and M*Y product feature information. The target graph model includes: constructing an adjacency matrix based on the nodes corresponding to the financial products in the M financial products, a group of connected adjacent nodes, and a group of connected edges, and constructing a feature matrix based on the M*Y product feature information; calculating the degree of each node in the adjacency matrix, and constructing a degree matrix based on the degree of each node; and integrating the adjacency matrix, feature matrix, and degree matrix to obtain the target graph model.

[0013] Furthermore, the initial product screening model based on the target graph model includes: aggregating information from the adjacency matrix, degree matrix, and feature matrix in the target graph model to obtain an aggregated matrix; obtaining a preset activation function, performing a linear transformation on the aggregated matrix using the preset activation function to obtain a transformation matrix, and adjusting the transformation matrix to obtain a first processing matrix; performing secondary information aggregation on the first processing matrix based on the preset activation function to obtain a second processing matrix, and processing the second processing matrix using a preset mapping function to obtain the initial product screening model.

[0014] To achieve the above objectives, according to another aspect of this application, a financial product screening device is provided. The device includes: an acquisition unit for acquiring user information of a financial user, determining product browsing data of the financial user over a historical time period based on the user information, and obtaining historical product browsing data; an input unit for inputting the historical product browsing data into a product screening model and outputting a list of recommended financial products, wherein the product screening model is trained on a graph convolutional neural network model using a training set, the graph convolutional neural network model is constructed from a target graph model, the training set consists of historical product browsing records of multiple financial users and the screened financial products associated with each financial user, and the financial product recommendation list includes multiple candidate financial products and a predicted score associated with each candidate financial product; and a screening unit for screening a preset number of candidate financial products from the financial product recommendation list to obtain target financial products, and sending the target financial products to the client used by the financial user.

[0015] According to another aspect of the present invention, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored executable program, wherein, when the executable program is executed, it controls the device where the computer-readable storage medium is located to perform the screening method for any of the above-mentioned financial products.

[0016] According to another aspect of the present invention, an electronic device is also provided, including one or more processors and a memory, the memory storing an executable program, and the processor for running the program, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement any of the above-described financial product screening methods.

[0017] According to another aspect of the present invention, a computer program product is also provided, the computer program product including a computer program, wherein when the computer program is executed by a processor, it implements the screening method for any of the above-mentioned financial products.

[0018] In this embodiment, a financial product filtering method is adopted. This involves acquiring the user information of financial users, determining their product browsing data over historical time periods based on this information, and obtaining historical product browsing data. This historical product browsing data is then input into a product filtering model, which outputs a list of recommended financial products. The product filtering model is trained on a graph convolutional neural network model using a training set. The graph convolutional neural network model is constructed from a target graph model. The training set consists of the historical product browsing records of multiple financial users and the filtered financial products associated with each user. The financial product recommendation list includes multiple candidate financial products and a predicted score associated with each candidate product. A preset number of candidate financial products are selected from the recommendation list to obtain the target financial product, which is then sent to the client used by the financial user. This method solves the technical problem of low filtering accuracy in related technologies. By acquiring the historical product browsing data of financial users, inputting this data into the product filtering model, outputting a list of recommended financial products, filtering the target financial product from the list, and sending the target financial product to the client used by the financial user, the technical effect of improving the filtering accuracy of financial products is achieved. Attached Figure Description

[0019] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0020] Figure 1 This is a hardware structure block diagram of a computer terminal (or mobile device) used to implement a screening method for financial products;

[0021] Figure 2 This is a flowchart of a financial product screening method provided according to an embodiment of this application;

[0022] Figure 3 This is a schematic diagram of a graph convolutional neural network model provided according to an embodiment of this application;

[0023] Figure 4 This is a schematic diagram of an optional financial product screening method provided according to an embodiment of this application;

[0024] Figure 5 This is a schematic diagram of a financial product screening device provided according to an embodiment of this application;

[0025] Figure 6 This is a structural block diagram of an electronic device according to an embodiment of this application. Detailed Implementation

[0026] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0027] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application 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 this application 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 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.

[0028] 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 application are information and data authorized by the user or fully authorized by all parties. For example, if there is an interface between this system and the relevant user or organization, before obtaining relevant information, it is necessary to send an acquisition request to the aforementioned user or organization through the interface, and obtain the relevant information after receiving consent information from the aforementioned user or organization.

[0029] It should be noted that the information collected in this application is information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of the relevant data all comply with the relevant laws, regulations and standards of the relevant regions, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation access points for users to choose to authorize use or refuse use.

[0030] Example 1

[0031] According to an embodiment of this application, a method embodiment for screening financial products is also 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. 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] The method embodiment provided in Embodiment 1 of this application can be executed on a mobile terminal, computer terminal, or similar computing device. Figure 1 This is a hardware structure block diagram of a computer terminal (or mobile device) used to implement a method for screening financial products, such as... Figure 1 As shown, computer terminal 10 (or mobile device) may include one or more ( Figure 1 (Illustrated as 102a, 102b, ..., 102n) Processor 102 (processor 102 may include, but is not limited to, a microprocessor MCU (Microcontroller Unit) or a programmable gate array (FPGA), etc.), memory 104 for storing data, and transmission device 106 for communication functions. In addition, it may include: a display, input / output interface (I / O interface), universal serial bus (USB) port (which may be included as one of the ports of a BUS bus), network interface, keyboard, cursor control device, power supply and / or camera. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the aforementioned electronic device. For example, computer terminal 10 may also include... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0033] It should be noted that the aforementioned one or more processors 102 and / or other data processing circuits are generally referred to herein as "data processing circuits". These data processing circuits may be embodied, in whole or in part, in software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuits may be a single, independent processing module, or may be integrated, in whole or in part, into any other element within the computer terminal 10 (or mobile device). As involved in the embodiments of this application, the data processing circuits serve as a processor control mechanism (e.g., selection of a variable resistor termination path connected to an interface).

[0034] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the financial product screening method in this embodiment. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, thereby realizing the aforementioned financial product screening method. The memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to the computer terminal 10 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0035] The transmission device 106 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the communication provider of the computer terminal 10. In one example, the transmission device 106 includes a network interface controller (NIC) and a network interface, which can be connected to other network devices via a base station to communicate with the Internet. In another example, the transmission device 106 may be a radio frequency (RF) module, used for wireless communication with the Internet.

[0036] The display can be, for example, a touchscreen liquid crystal display (LCD), which allows the user to interact with the user interface of the computer terminal 10 (or mobile device).

[0037] Under the aforementioned operating environment, this application provides the following: Figure 2 The method for screening financial products is shown. Figure 2 This is a flowchart of a financial product screening method provided according to an embodiment of this application, such as... Figure 2 As shown, the method includes the following steps:

[0038] Step S201: Obtain the user information of the financial user, determine the product browsing data of the financial user in the historical time period based on the user information, and obtain the historical product browsing data.

[0039] Specifically, to overcome the current reliance on excessive user historical behavior information and the resulting waterfall effect (where recommendations concentrate on a few popular financial products while ignoring less popular ones), the first step is to obtain user information of the financial users whose products are to be recommended. This information could include basic information, account information, and historical behavior data. Then, historical behavior data is extracted from this user information. Based on this data, analysis of user browsing behavior at different times is needed to identify seasonal or cyclical investment preferences, thereby determining the financial user's product browsing data over historical periods—that is, historical product browsing data. This historical behavior data can include user browsing history on financial platforms, search keywords, favorites, purchase records of financial products, and investment preferences.

[0040] Step S202: Input historical product browsing data into the product filtering model and output a list of recommended financial products. The product filtering model is trained on a graph convolutional neural network model using a training set. The graph convolutional neural network model is constructed from a target graph model. The training set consists of the historical product browsing records of multiple financial users and the filtered financial products associated with each financial user. The list of recommended financial products includes multiple candidate financial products and the predicted score associated with each candidate financial product.

[0041] Specifically, after obtaining the historical product browsing data of financial users, the historical product browsing data can first be input into a product filtering model constructed from the historical product browsing records of multiple financial users. Then, the model outputs a financial product recommendation list containing multiple candidate financial products. The candidate financial products included in the financial product recommendation list are a list constructed based on the predicted scores. The product filtering model can refer to a graph convolutional neural network model, which is constructed from a target graph model. In the target graph model, the nodes represent users and financial products, and the edges represent the interaction relationship between users and products, such as browsing, purchasing, and consulting.

[0042] It should be noted that during training, graph convolutional neural network models can update the feature vector of each node through multiple graph convolutional layers. This process integrates the neighbor information of the node, learns deeper node feature representations through weight sharing and activation functions, and aims to minimize the gap between the predicted rating and the actual user rating during training. Cross-entropy loss or other similarity metrics are used to evaluate the model performance.

[0043] Step S203: Select a preset number of candidate financial products from the financial product recommendation list to obtain the target financial product, and send the target financial product to the client used by the financial user.

[0044] Specifically, after obtaining the list of recommended financial products, multiple candidate financial products with high prediction scores can be selected from the list, and then these candidate financial products can be sent to financial users as target financial products. It should be noted that in order to avoid the waterfall effect, a diversification strategy can be adopted to ensure that the recommendation list contains different types of financial products.

[0045] Furthermore, after being sent to financial users, the system can continuously monitor user behavior and adjust the parameters of the graph convolutional neural network, such as the number of iterations, learning rate, and node feature dimensions, based on user feedback, in order to optimize the performance of the recommendation algorithm.

[0046] The financial product screening method provided in this application embodiment obtains the user information of financial users, determines the product browsing data of financial users in historical time periods based on the user information, and obtains historical product browsing data; inputs the historical product browsing data into a product screening model, and outputs a financial product recommendation list. The product screening model is trained on a graph convolutional neural network model using a training set, and the graph convolutional neural network model is constructed from a target graph model. The training set consists of the historical product browsing records of multiple financial users and the screened financial products associated with each financial user. The financial product recommendation list includes multiple candidate financial products and a predicted score associated with each candidate financial product; a preset number of candidate financial products are selected from the financial product recommendation list to obtain the target financial product, and the target financial product is sent to the client used by the financial user. This method solves the technical problem of low screening accuracy in related technologies. By obtaining the historical product browsing data of financial users, inputting the historical product browsing data into the product screening model, outputting a financial product recommendation list, selecting the target financial product from the financial product recommendation list, and sending the target financial product to the client used by the financial user, the technical effect of improving the screening accuracy of financial products is achieved.

[0047] Optionally, in the financial product screening method provided in this application embodiment, the product screening model is trained in the following manner: A list of financial products is obtained, and browsing records associated with M financial products in the list are obtained, resulting in M ​​sets of browsing records, where M is a positive integer; for each financial product, Y financial products associated with the financial product are selected from the set of browsing records associated with each financial product, resulting in M ​​sets of associated financial products, where Y is a positive integer; a graph model is constructed based on the M sets of associated financial products and the M financial products to obtain a target graph model, and an initial product screening model is built based on the target graph model; historical product browsing records of K financial users and the screened financial products associated with each financial user are obtained, and a training set is constructed from the K historical product browsing records and the K screened financial products, where K is a positive integer; the initial product screening model is trained using the training set to obtain the product screening model.

[0048] Specifically, before inputting historical product browsing data into the product selection model, it is necessary to train against blind browsing. First, a list of financial products containing multiple categories can be obtained, and each product can be numbered. For example, the list could include fixed deposits, bonds, money market funds, insurance, and stocks. Simultaneously, the browsing history of financial users who purchased each product from the aforementioned list can be obtained, resulting in multiple sets of browsing records.

[0049] Furthermore, for each financial product, the browsing history associated with that product is used to filter out the financial products with the highest correlation. For example, if 100 users purchased product A, the names of the 10 financial products browsed by each user before purchasing product A are obtained, resulting in 1000 browsing records. The 6 products with the highest frequency are then selected from these 1000 records and linked to product A, thus obtaining a set of associated financial products. A graph model is then constructed using these associated financial products and the financial products in the product list. This model not only captures the relationships between products but also incorporates the characteristics of the products themselves, such as category, risk level, popularity, and release time, ultimately yielding the target graph model.

[0050] Furthermore, a graph convolutional neural network model can be constructed using the aforementioned target graph model. The model can then be trained using a training set obtained from multiple historical product browsing records and corresponding filtered financial products. This method yields a product selection model. The training process may include adjusting model parameters and optimizing model performance to minimize the difference between predictions and actual purchasing behavior.

[0051] This embodiment trains a product filtering model by utilizing users' historical browsing data and the correlation between products. This enables the model to filter out products that users may be interested in, improving the diversity and comprehensiveness of recommendations. Furthermore, it can update the recommendation list in real time as user behavior changes, ensuring the up-to-dateness and relevance of the recommendation results, adapting to users' immediate needs, and enhancing the user experience.

[0052] Optionally, in the financial product filtering method provided in this application embodiment, obtaining the browsing records associated with M financial products in the list of purchased financial products to obtain M sets of browsing records includes: obtaining the financial users of each financial product in the list of purchased financial products to obtain M sets of financial users, wherein each set of financial users includes N financial users, where N is a positive integer; for each financial product, obtaining the browsing records of a set of financial users corresponding to each financial product before purchasing each financial product to obtain M sets of browsing records, wherein each set of browsing records includes R browsing records, where R is a positive integer.

[0053] Specifically, when obtaining browsing history, the process begins by retrieving a list of buyers for each financial product from the financial institution's transaction database. This list is then used to identify the financial users for each product. Next, for each financial product, the browsing history of a corresponding group of financial users before purchasing that product can be collected, thus obtaining the relevant browsing records.

[0054] Furthermore, after obtaining multiple sets of browsing records, a graph model can be constructed using these records. In this model, each user and each financial product serves as a node, while the edges reflect the user's browsing behavior towards the product, including browsing frequency, browsing time, and browsing depth. Through the graph model, not only can the user's attention to a specific financial product be displayed intuitively, but the similarity between users and the correlation between products can also be revealed, providing rich structured data for the subsequent training of the product selection model.

[0055] This embodiment obtains the browsing history of financial users purchasing financial products, and then constructs a graph model based on the browsing history to lay a data foundation for subsequent screening of financial products.

[0056] Optionally, in the financial product screening method provided in this application embodiment, screening Y financial products associated with a financial product from a set of browsing records associated with each financial product includes: for a financial product, screening Q browsed financial products from each browsing record associated with the financial product to obtain R sets of browsed financial products, where Q is a positive integer; determining the frequency data of each browsed financial product in the R sets of browsed financial products, screening the R sets of browsed financial products based on the frequency data of each browsed financial product to obtain Y browsed financial products with the highest frequency data; and determining the Y browsed financial products with the highest frequency data as Y financial products.

[0057] Specifically, in order to filter the associated financial products for each financial product, when a user browses a certain financial product, for each main financial product browsed, multiple financial products that the user viewed before browsing this product can be filtered from each of its associated browsing records, thus obtaining multiple sets of browsed financial products.

[0058] Furthermore, the frequency data of each browsed financial product can be determined, and the browsed financial products can be sorted according to the frequency data. The browsed financial products with the highest frequency data are then selected as associated financial products. For example, if 100 users have purchased product A, the 10 most recent browsing records of each user before purchasing product A can be obtained, resulting in a total of 1,000 browsing records. Each record includes a browsed financial product. The 6 products with the highest frequency are then selected from the 1,000 records, and these 6 financial products are linked to product A to obtain a set of associated financial products.

[0059] This embodiment analyzes user browsing behavior to identify associated financial products for each financial product. The resulting graph model can more accurately capture user interests and needs, improve user browsing experience and satisfaction on the platform, and significantly increase conversion rates.

[0060] Optionally, in the financial product screening method provided in this application embodiment, constructing a graph model based on M groups of related financial products and M financial products to obtain the target graph model includes: for the M financial products, obtaining product feature information for each financial product to obtain M*Y product feature information, wherein each product feature information includes at least one of the following: product type, risk level, screening popularity, and release time; determining the financial product as a node, determining a group of related financial products corresponding to the financial product as adjacent nodes of the node, and determining the association relationship between the financial product and each related financial product as an edge; constructing a graph model based on the M nodes, M groups of adjacent nodes, M groups of edges, and M*Y product feature information to obtain the target graph model.

[0061] Specifically, when constructing a graph model, it is first necessary to collect detailed feature information for each financial product, such as product type, risk level, screening popularity, and release time. The product feature information is denoted as feature, which includes product type (class), risk level (riskLever), screening popularity (hot), and release time (releaseTime), i.e., feature = (class, riskLever, hot, releaseTime). Product type includes three-dimensional features, risk level includes four-dimensional features, screening popularity includes four-dimensional features, and release time includes five-dimensional features. At this time, the product feature information includes 16-dimensional features. It should be noted that when the product type is a fixed deposit, the feature can be 000; when the product type is a bond, the feature can be 001; when the product type is a money market fund, the feature can be 010, etc.; the popularity level is divided into 16 levels, such as the feature for popularity level 0 being 0000, the feature for popularity level 1 being 0001, and the feature for popularity level 15 being 1111; the risk level is divided into 16 levels, with higher values ​​indicating higher risk, such as the feature for risk level 0 being 0000, and the feature for risk level 15 being 1111; the release time is the number of days since the product was released, with newly released products having smaller feature values.

[0062] Furthermore, a graph model is constructed around financial products, where each financial product is represented as a node, a set of related financial products corresponding to a financial product is identified as the adjacent nodes of the node, and the relationships between a financial product and each related financial product are defined as edges. Based on these nodes and edges, the target graph model is constructed. Each node stores various information such as product type, risk level, popularity, and release time. For example, if a user browses product A and then frequently views products B and C, then products B and C can be considered adjacent nodes of product A.

[0063] This embodiment constructs a target graph model using financial products and related financial products. This target graph model can capture the implicit relationships between products and reflect user behavior and product status, laying a data foundation for improving the accuracy of recommendations and user experience.

[0064] Optionally, in the financial product screening method provided in this application embodiment, constructing a graph model based on M nodes, M groups of adjacent nodes, M groups of edges, and M*Y product feature information to obtain the target graph model includes: constructing an adjacency matrix based on the nodes corresponding to the financial products in the M financial products, a group of connected adjacent nodes, and a group of connected edges, and constructing a feature matrix based on the M*Y product feature information; calculating the degree of each node in the adjacency matrix, and constructing a degree matrix based on the degree of each node; and integrating the adjacency matrix, feature matrix, and degree matrix to obtain the target graph model.

[0065] Specifically, an adjacency matrix is ​​a mathematical matrix that describes the connections between nodes. When constructing a graph model, we can first define each product as a node and the relationships between products (defined by user browsing behavior) as edges to build an adjacency matrix. In the adjacency matrix, each row and column represents a node, and the element value (0 or 1) indicates whether there is an edge connecting the nodes. If there is a connection, the value is 1; otherwise, it is 0. For example, if a user browses product i and then browses product j, the value of row i and column j in the adjacency matrix is ​​1, indicating that there is a connection between i and j, i.e., A[i][j] = 1. It should be noted that for each pair of financial products, in order to more accurately reflect the strength of the connection between different products, a weight can be set, or the weight of the edge can be adjusted based on the frequency of user browsing behavior or other relevant indicators.

[0066] Furthermore, a feature matrix is ​​constructed using the product feature information corresponding to each node. Each row in the matrix corresponds to a product node, and each column represents a product feature (such as product type, risk level, popularity, release time, etc.). The features of each financial product consist of 16 dimensions. Then, the degree of each node in the adjacency matrix is ​​calculated, which is equivalent to calculating the number of non-zero elements in each row of the adjacency matrix. This yields the degree of each node, ensuring that the diagonal element D[i][i] of the degree matrix equals the degree of the i-th node, i.e., the number of non-zero elements in the i-th row of the adjacency matrix. Finally, after constructing the adjacency matrix, feature matrix, and degree matrix, these matrices are integrated to obtain the target graph model.

[0067] This embodiment integrates the adjacency matrix, feature matrix, and degree matrix to obtain the target graph model. The graph model can not only capture the direct relationship between products, but also learn indirect relationships and deeper feature relationships through iterative graph convolution operations, generating more personalized and accurate product recommendations and significantly improving the user experience.

[0068] Optionally, in the financial product screening method provided in this application embodiment, building an initial product screening model based on a target graph model includes: aggregating information from the adjacency matrix, degree matrix, and feature matrix in the target graph model to obtain an aggregated matrix; obtaining a preset activation function, performing a linear transformation on the aggregated matrix using the preset activation function to obtain a transformation matrix, and adjusting the transformation matrix to obtain a first processing matrix; performing secondary information aggregation on the first processing matrix based on the preset activation function to obtain a second processing matrix, and processing the second processing matrix using a preset mapping function to obtain the initial product screening model.

[0069] Specifically, in order to construct a Graph Convolutional Network (GCN) model, and then obtain an initial product screening model based on the GCN model, Figure 3 This is a schematic diagram of a graph convolutional neural network model provided according to an embodiment of this application, such as... Figure 3 As shown, the adjacency matrix A, degree matrix D, and feature matrix H of the graph model are first aggregated in the first convolutional layer (i.e., GCN Layer 1). The aggregated matrix can be obtained by the following formula:

[0070]

[0071] Where A is the adjacency matrix of the graph model, D is the degree matrix corresponding to the adjacency matrix, H is the feature matrix of the graph model, W is the convolution kernel of the l-th layer, and σ is the activation function, which can be ReLU. The adjacency matrix describes the connections between nodes; the degree matrix records the number of connections or "degree" for each node; and the feature matrix contains the feature information of each node (i.e., each financial product). Through information aggregation, the feature information of each node can be updated, and the information of neighboring nodes can be integrated, which helps to capture the complex relationships between financial products.

[0072] Furthermore, the aggregated information undergoes a linear transformation using the weight matrix, followed by a pre-defined activation function (e.g., ReLU) to enhance the model's non-linear expressive power and better capture the relationships between complex features, thus obtaining the transformation matrix. Since in the original adjacency matrix A, A[i,j] = 1 indicates a connection between node i and node j, and otherwise A[i,j] = 0, directly using the adjacency matrix for graph convolution may restrict information propagation at some nodes, failing to effectively integrate information from other nodes in the graph. To overcome this problem, a self-loop can be added to the adjacency matrix A, i.e., A[i,i] = 1, with diagonal elements set to 1. This ensures that information propagation in the graph convolutional layer is not interrupted due to the lack of neighbors, effectively normalizing the transformation matrix to obtain the first processing matrix. This normalization not only ensures that all nodes participate in information propagation but also ensures that information propagation in the graph is not affected by node degree. Each node's contribution to its neighboring nodes' information updates is balanced, contributing to improved model stability.

[0073] Furthermore, after the first round of graph convolution, the node features are processed by the ReLU activation function. This helps increase the model's non-linearity, enabling it to learn more complex feature patterns. At this point, information aggregation can be performed again in the second convolutional layer (GCN Layer 2). Based on a preset activation function, the first processing matrix is ​​aggregated a second time to ensure the final node features have sufficient non-linearity. The node features are then further fused to obtain the second processing matrix. Finally, the node features are passed through a classification layer (such as the Softmax mapping function) to map them to probability distributions for different categories. In other words, the second processing matrix is ​​processed using a preset mapping function, thus obtaining the initial product selection model.

[0074] This embodiment uses multi-layer information aggregation and nonlinear transformation to enable the graph convolutional neural network model to learn complex and deep feature representations, capture the correlation between financial products and the dynamic changes in user interests, and then, based on the updated features of nodes, the initial product screening model can generate personalized product recommendations for each user, improving the accuracy of recommendations and user satisfaction.

[0075] This application also provides a method for screening financial products. Figure 4 This is a schematic diagram of an optional financial product screening method provided according to an embodiment of this application, such as... Figure 4 As shown, the method includes:

[0076] To recommend wealth management products within the financial product category, one can first collect information on all wealth management products (including fixed deposits, government bonds, money market funds, insurance, and stocks), construct a wealth management product dataset based on product characteristics, and then perform a dataset partitioning operation to divide the dataset into a training set and a test set.

[0077] Furthermore, for each wealth management product in the training set, 16-dimensional feature information, including product type (class), risk level (riskLever), popularity (hot), and release time (releaseTime), can be obtained. This means the 16-dimensional features of the training samples are obtained. For example, a fixed deposit product has the following features: class = 000, riskLever = 0000, hot = 0010, releaseTime = 00001. Then, based on the link information when a user browses and purchases wealth management products, a graph model of the wealth management products is constructed. In this graph model, products are treated as nodes, and related products before and after a user's browsing are treated as adjacent nodes. The weight of each edge is set to 1 (indicating that the user browses from one product to another). For example, if a user browses product A and then most frequently browses products B, C, D, E, F, and G, these products can be treated as nodes, and the relationships between products can be treated as edges in the graph model.

[0078] Furthermore, a two-layer graph convolutional neural network model is built based on the aforementioned graph model of financial products using a deep learning framework. The model's input is the feature matrix from the graph model, and its output is the financial product recommendation result. The model is then trained and tested using a divided training and test set. The number of epochs can be set to 600, and the learning rate (lr) to 0.00005. The network is trained using the training set to optimize network parameters and reduce the loss value output by the loss function. After obtaining the final model, it is tested on the test set to improve the accuracy of the model's financial product recommendations. After training, test samples can be input into the network to output recommended financial product structures. For example, if a user recently viewed fixed deposit products, the model can recommend low-risk and popular money market funds to meet the user's potential needs.

[0079] This embodiment constructs a graph model of financial products by using purchase click records of financial products as edges and building features for each node based on the user's browsing history. Then, it uses the graph model to generate a graph convolutional neural network model, which can recommend personalized financial products in real time based on the user's browsing behavior and historical preferences, significantly improving the accuracy and satisfaction of the recommendations and increasing the response speed.

[0080] 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, and 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.

[0081] Example 2

[0082] This application also provides a financial product screening device. It should be noted that the financial product screening device of this application can be used to execute the financial product screening method provided in this application. The following describes the financial product screening device provided in this application.

[0083] According to an embodiment of this application, an apparatus for implementing the above-described financial product screening method is also provided. Figure 5 This is a schematic diagram of a financial product screening device provided according to an embodiment of this application, such as... Figure 5 As shown, the device includes: an acquisition unit 50, an input unit 51, and a filtering unit 52.

[0084] The acquisition unit 50 is used to acquire user information of financial users, determine product browsing data of financial users in historical time periods based on user information, and obtain historical product browsing data.

[0085] Input unit 51 is used to input historical product browsing data into the product filtering model and output a list of recommended financial products. The product filtering model is trained on a graph convolutional neural network model using a training set. The graph convolutional neural network model is constructed from a target graph model. The training set consists of the historical product browsing records of multiple financial users and the filtered financial products associated with each financial user. The list of recommended financial products includes multiple candidate financial products and the predicted score associated with each candidate financial product.

[0086] The filtering unit 52 is used to filter out a preset number of candidate financial products from the financial product recommendation list, obtain the target financial product, and send the target financial product to the client used by the financial user.

[0087] The financial product filtering device provided in this application embodiment acquires user information of financial users through an acquisition unit 50, determines the product browsing data of financial users in historical time periods based on the user information, and obtains historical product browsing data; an input unit 51 inputs the historical product browsing data into a product filtering model and outputs a financial product recommendation list. The product filtering model is trained on a graph convolutional neural network model using a training set, and the graph convolutional neural network model is constructed from a target graph model. The training set consists of the historical product browsing records of multiple financial users and the filtered financial products associated with each financial user. The financial product recommendation list includes multiple candidate financial products and the predicted score associated with each candidate financial product; a filtering unit 52 filters a preset number of candidate financial products from the financial product recommendation list to obtain target financial products, and sends the target financial products to the client used by the financial user. This solves the technical problem of low filtering accuracy in related technologies. By acquiring the historical product browsing data of financial users, inputting the historical product browsing data into the product filtering model, outputting a financial product recommendation list, filtering the target financial products from the financial product recommendation list, and sending the target financial products to the client used by the financial user, the technical effect of improving the filtering accuracy when filtering financial products is achieved.

[0088] Optionally, in the financial product screening device provided in this application embodiment, the input unit 51 includes: a first acquisition module, used to acquire a list of financial products and acquire browsing records associated with M financial products in the list of purchased financial products, to obtain M sets of browsing records, where M is a positive integer; a first screening module, used to, for each financial product, select Y financial products associated with the financial product from a set of browsing records associated with each financial product, to obtain M sets of associated financial products, where Y is a positive integer; a first construction module, used to construct a graph model based on the M sets of associated financial products and the M financial products, to obtain a target graph model, and to build an initial product screening model based on the target graph model; a second acquisition module, used to acquire historical product browsing records of K financial users and the screened financial products associated with each financial user, to construct a training set from the K historical product browsing records and the K screened financial products, where K is a positive integer; and a training module, used to train the initial product screening model using the training set to obtain a product screening model.

[0089] Optionally, in the financial product screening device provided in this application embodiment, the input unit 51 includes: a third acquisition module, used to acquire financial users who purchase each financial product in the list of financial products, to obtain M groups of financial users, wherein each group of financial users includes N financial users, and N is a positive integer; and a fourth acquisition module, used to acquire, for each financial product, the browsing records of a group of financial users corresponding to each financial product before purchasing each financial product, to obtain M groups of browsing records, wherein each group of browsing records includes R browsing records, and R is a positive integer.

[0090] Optionally, in the financial product screening device provided in this application embodiment, the input unit 51 includes: a second screening module, used to screen out Q browsed financial products from each browsing record associated with a financial product to obtain R groups of browsed financial products, where Q is a positive integer; a first determining module, used to determine the frequency data of each browsed financial product in the R groups of browsed financial products, and screen the R groups of browsed financial products based on the frequency data of each browsed financial product to obtain Y browsed financial products with the highest frequency data; and a second determining module, used to determine the Y browsed financial products with the highest frequency data as Y financial products.

[0091] Optionally, in the financial product screening device provided in this application embodiment, the input unit 51 includes: a fifth acquisition module, used to acquire product feature information of each of the M financial products to obtain M*Y product feature information, wherein each product feature information includes at least one of the following: product type, risk level, screening popularity, and release time; a third determination module, used to determine the financial product as a node, determine a group of related financial products corresponding to the financial product as adjacent nodes of the node, and determine the association relationship between the financial product and each related financial product as an edge; and a second construction module, used to construct a graph model based on the M nodes, M groups of adjacent nodes, M groups of edges, and M*Y product feature information to obtain a target graph model.

[0092] Optionally, in the financial product screening device provided in this application embodiment, the input unit 51 includes: a third construction module, used to construct an adjacency matrix based on the nodes corresponding to the financial products in M ​​financial products, a set of connected adjacent nodes, and a set of connected edges, and to construct a feature matrix based on the feature information of M*Y products; a calculation module, used to calculate the degree of each node in the adjacency matrix and construct a degree matrix based on the degree of each node; and an integration module, used to integrate the adjacency matrix, the feature matrix, and the degree matrix to obtain a target graph model.

[0093] Optionally, in the financial product screening device provided in this application embodiment, the input unit 51 includes: a first aggregation module, used to aggregate information from the adjacency matrix, degree matrix, and feature matrix in the target graph model to obtain an aggregated matrix; a sixth acquisition module, used to acquire a preset activation function, perform a linear transformation on the aggregated matrix using the preset activation function to obtain a transformation matrix, and adjust the transformation matrix to obtain a first processing matrix; and a second aggregation module, used to perform secondary information aggregation on the first processing matrix based on the preset activation function to obtain a second processing matrix, and process the second processing matrix using a preset mapping function to obtain an initial product screening model.

[0094] It should be noted that the acquisition unit 50, input unit 51, and filtering unit 52 mentioned above correspond to steps S201 to S203 in Embodiment 1. The instances and application scenarios implemented by the above units and corresponding steps are the same, but are not limited to the content disclosed in Embodiment 1. It should be noted that the above modules or units can be hardware or software components stored in memory (e.g., memory 104) and processed by one or more processors (e.g., processors 102a, 102b, ..., 102n). The above units can also be part of a device and run in the computer terminal 10 provided in Embodiment 1.

[0095] Example 3

[0096] Embodiments of this application may provide a computer terminal, which may be any computer terminal device in a group of computer terminals. Optionally, in this embodiment, the aforementioned computer terminal may also be replaced with a mobile terminal or an electronic device, etc.

[0097] Optionally, in this embodiment, the computer terminal may be located in at least one of a plurality of network devices in a computer network.

[0098] In this embodiment, the computer terminal described above can execute the program code for the following steps in the financial product screening method: obtaining user information of financial users, determining product browsing data of financial users in historical time periods based on user information, and obtaining historical product browsing data; inputting the historical product browsing data into a product screening model and outputting a financial product recommendation list, wherein the product screening model is obtained by training a graph convolutional neural network model on a training set, the graph convolutional neural network model is constructed from a target graph model, the training set consists of the historical product browsing records of multiple financial users and the screened financial products associated with each financial user, and the financial product recommendation list includes multiple candidate financial products and the predicted score associated with each candidate financial product; selecting a preset number of candidate financial products from the financial product recommendation list to obtain the target financial product, and sending the target financial product to the client used by the financial user.

[0099] Optionally, the aforementioned computer terminal can execute the program code for the following steps in the financial product screening method: obtaining a list of financial products and acquiring browsing records associated with M financial products in the list, resulting in M ​​sets of browsing records, where M is a positive integer; for each financial product, selecting Y financial products associated with it from the set of browsing records associated with each financial product, resulting in M ​​sets of associated financial products, where Y is a positive integer; constructing a graph model based on the M sets of associated financial products and the M financial products, resulting in a target graph model, and building an initial product screening model based on the target graph model; acquiring historical product browsing records of K financial users and the filtered financial products associated with each financial user, constructing a training set from the K historical product browsing records and the K filtered financial products, where K is a positive integer; training the initial product screening model using the training set, resulting in a product screening model.

[0100] Optionally, the computer terminal described above can execute the program code for the following steps in the financial product screening method: obtain the financial users who purchased each financial product from the list of financial products, resulting in M ​​groups of financial users, where each group of financial users includes N financial users, and N is a positive integer; for each financial product, obtain the browsing records of the corresponding group of financial users before purchasing each financial product, resulting in M ​​groups of browsing records, where each group of browsing records includes R browsing records, and R is a positive integer.

[0101] Optionally, the computer terminal described above can execute the program code for the following steps in the financial product screening method: For a financial product, select Q browsed financial products from each browsing record associated with the financial product to obtain R groups of browsed financial products, where Q is a positive integer; determine the frequency data of each browsed financial product in the R groups of browsed financial products, and screen the R groups of browsed financial products based on the frequency data of each browsed financial product to obtain Y browsed financial products with the highest frequency data; determine the Y browsed financial products with the highest frequency data as Y financial products.

[0102] Optionally, the aforementioned computer terminal can execute the program code for the following steps in the financial product screening method: For M financial products, obtain the product feature information of each financial product to obtain M*Y product feature information, wherein each product feature information includes at least one of the following: product type, risk level, screening popularity, and release time; determine the financial product as a node, determine a group of related financial products corresponding to the financial product as adjacent nodes of the node, and determine the relationship between the financial product and each related financial product as an edge; construct a graph model based on the M nodes, M groups of adjacent nodes, M groups of edges, and M*Y product feature information to obtain the target graph model.

[0103] Optionally, the computer terminal described above can execute the program code for the following steps in the financial product screening method: constructing an adjacency matrix based on the nodes corresponding to the financial products in the M financial products, a set of connected adjacent nodes, and a set of connected edges, and constructing a feature matrix based on the feature information of the M*Y products; calculating the degree of each node in the adjacency matrix, and constructing a degree matrix based on the degree of each node; integrating the adjacency matrix, feature matrix, and degree matrix to obtain the target graph model.

[0104] Optionally, the aforementioned computer terminal can execute the program code for the following steps in the financial product screening method: aggregate information from the adjacency matrix, degree matrix, and feature matrix in the target graph model to obtain an aggregated matrix; obtain a preset activation function, perform a linear transformation on the aggregated matrix using the preset activation function to obtain a transformation matrix, and adjust the transformation matrix to obtain a first processing matrix; perform secondary information aggregation on the first processing matrix based on the preset activation function to obtain a second processing matrix, and process the second processing matrix using a preset mapping function to obtain an initial product screening model.

[0105] Optionally, Figure 6 This is a structural block diagram of an electronic device according to an embodiment of this application. Figure 6 As shown, the electronic device may include: one or more ( Figure 6 (Only one is shown) processor 602, memory 604, memory controller, and peripheral interface, wherein the peripheral interface is connected to the radio frequency module, audio module and display.

[0106] The memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the financial product screening method and apparatus in this application embodiment. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, thereby realizing the aforementioned financial product screening method. The memory may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include memory remotely located relative to the processor, and these remote memories can be connected to the terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0107] The processor can access the information and application programs stored in the memory via the transmission device to execute the steps described above in the financial product screening method.

[0108] This application provides a scheme for screening financial products. By acquiring user information of financial users, and determining their product browsing data over historical time periods based on this information, historical product browsing data is obtained. This historical product browsing data is then input into a product screening model, which outputs a list of recommended financial products. The product screening model is trained on a graph convolutional neural network model using a training set. The graph convolutional neural network model is constructed from a target graph model. The training set consists of the historical product browsing records of multiple financial users and the screened financial products associated with each user. The recommended financial product list includes multiple candidate financial products and a predicted score associated with each candidate product. A preset number of candidate financial products are selected from the recommended financial product list to obtain target financial products, which are then sent to the client used by the financial user. This solves the technical problem of low screening accuracy in related technologies. By acquiring historical product browsing data of financial users, inputting this data into the product screening model, outputting a list of recommended financial products, selecting target financial products from the list, and sending the target financial products to the client used by the financial user, the technical effect of improving the screening accuracy of financial products is achieved.

[0109] Those skilled in the art will understand that Figure 6 The structure shown is for illustrative purposes only. Electronic devices can also be smartphones, tablets, handheld computers, mobile internet devices (MIDs), PADs, and other terminal devices. Figure 6 This does not limit the structure of the aforementioned electronic device. For example, electronic devices may also include components that are more... Figure 6The more or fewer components shown (such as network interfaces, display devices, etc.), or having the same Figure 6 The different configurations shown.

[0110] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing the hardware related to the terminal device. The program can be stored in a computer-readable storage medium, which may include: flash drive, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0111] Example 4

[0112] Embodiments of this application also provide a storage medium. Optionally, in this embodiment, the storage medium can be used to store the program code executed by the financial product screening method provided in Embodiment 1.

[0113] Optionally, in this embodiment, the storage medium may be located in any computer terminal in a group of computer terminals in a computer network, or in any mobile terminal in a group of mobile terminals.

[0114] Optionally, in this embodiment, the storage medium is configured to store program code for performing the following steps: obtaining user information of financial users, determining product browsing data of financial users in historical time periods based on user information, and obtaining historical product browsing data; inputting historical product browsing data into a product filtering model and outputting a list of recommended financial products, wherein the product filtering model is trained on a graph convolutional neural network model using a training set, the graph convolutional neural network model is constructed from a target graph model, the training set consists of historical product browsing records of multiple financial users and filtered financial products associated with each financial user, and the list of recommended financial products includes multiple candidate financial products and a predicted score associated with each candidate financial product; filtering a preset number of candidate financial products from the list of recommended financial products to obtain target financial products, and sending the target financial products to the client used by the financial user.

[0115] This application also provides a computer program product that, when executed on a data processing device, is suitable for performing the steps of a screening method for financial products.

[0116] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0117] In the above embodiments of this application, 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.

[0118] 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 is only 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 coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection of units or modules may be electrical or other forms.

[0119] 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 network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0120] Furthermore, the functional units in the various embodiments of this application 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.

[0121] 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 this application, 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 this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0122] The above are merely preferred embodiments of this application. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method of screening for a financial product, characterized by, include: Obtain user information of financial users, and determine the product browsing data of financial users in historical time periods based on the user information to obtain historical product browsing data; The historical product browsing data is input into the product filtering model, and a list of recommended financial products is output. The product filtering model is trained on a graph convolutional neural network model using a training set. The graph convolutional neural network model is constructed from a target graph model. The training set consists of the historical product browsing records of multiple financial users and the filtered financial products associated with each financial user. The list of recommended financial products includes multiple candidate financial products and the predicted score associated with each candidate financial product. A preset number of candidate financial products are selected from the financial product recommendation list to obtain the target financial product, which is then sent to the client used by the financial user.

2. The method according to claim 1, characterized in that, The product selection model was trained in the following way: Obtain a list of financial products, and obtain the browsing records associated with purchasing M financial products in the list, resulting in M ​​sets of browsing records, where M is a positive integer; For each financial product, select Y financial products associated with the financial product from a set of browsing records associated with each financial product to obtain M sets of associated financial products, where Y is a positive integer; Based on the M groups of related financial products and the M financial products, a graph model is constructed to obtain the target graph model, and an initial product screening model is built based on the target graph model. Obtain the historical product browsing records of K financial users and the filtered financial products associated with each financial user. Construct the training set from the K historical product browsing records and the K filtered financial products, where K is a positive integer. The initial product selection model is trained using the training set to obtain the product selection model.

3. The method according to claim 2, characterized in that, Retrieve the browsing history associated with M financial products in the list of financial products, resulting in M ​​sets of browsing history, including: Obtain the financial users who have purchased each financial product in the list of financial products, resulting in M ​​groups of financial users, where each group of financial users includes N financial users, and N is a positive integer; For each financial product, obtain the browsing records of a group of financial users before purchasing each financial product, resulting in the M groups of browsing records, where each group of browsing records includes R browsing records, and R is a positive integer.

4. The method according to claim 2, characterized in that, From a set of browsing records associated with each financial product, select Y financial products associated with the stated financial product, including: For a financial product, select Q browsed financial products from each browsing record associated with the financial product to obtain R groups of browsed financial products, where Q is a positive integer; Determine the frequency data of each browsed financial product in the R groups of browsed financial products, and filter the R groups of browsed financial products based on the frequency data of each browsed financial product to obtain Y browsed financial products with the highest frequency data. The financial products with the highest frequency of browsing among the Y data points are identified as the Y financial products.

5. The method according to claim 2, characterized in that, Based on the M groups of related financial products and the M individual financial products, a graph model is constructed, resulting in the target graph model, which includes: For the M financial products, product feature information for each financial product is obtained, resulting in M*Y product feature information. Each product feature information includes at least one of the following: product type, risk level, screening popularity, and release time. The financial product is defined as a node, a group of associated financial products corresponding to the financial product is defined as the adjacent nodes of the node, and the relationship between the financial product and each associated financial product is defined as an edge. The target graph model is obtained by constructing a graph model based on M nodes, M groups of adjacent nodes, M groups of edges, and M*Y product feature information.

6. The method according to claim 5, characterized in that, Based on M nodes, M groups of adjacent nodes, M groups of edges, and M*Y product feature information, a graph model is constructed, resulting in the target graph model, which includes: An adjacency matrix is ​​constructed based on the nodes corresponding to the financial products among the M financial products, a set of adjacent nodes connected to each other, and a set of edges connected to each other. A feature matrix is ​​constructed based on the feature information of the M*Y products. Calculate the degree of each node in the adjacency matrix, and construct a degree matrix based on the degree of each node; The adjacency matrix, the feature matrix, and the degree matrix are integrated to obtain the target graph model.

7. The method according to claim 2, characterized in that, The initial product screening model built based on the target graph model includes: Information aggregation is performed on the adjacency matrix, degree matrix, and feature matrix in the target graph model to obtain the aggregated matrix; Obtain a preset activation function, perform a linear transformation on the aggregated matrix using the preset activation function to obtain a transformation matrix, and adjust the transformation matrix to obtain a first processing matrix; The first processing matrix is ​​subjected to secondary information aggregation based on the preset activation function to obtain the second processing matrix. The second processing matrix is ​​then processed using a preset mapping function to obtain the initial product screening model.

8. A screening device for financial products, characterized in that, include: The acquisition unit is used to acquire user information of financial users, determine product browsing data of financial users in historical time periods based on the user information, and obtain historical product browsing data. The input unit is used to input the historical product browsing data into the product filtering model and output a list of recommended financial products. The product filtering model is trained on a graph convolutional neural network model using a training set. The graph convolutional neural network model is constructed from a target graph model. The training set consists of the historical product browsing records of multiple financial users and the filtered financial products associated with each financial user. The list of recommended financial products includes multiple candidate financial products and the predicted score associated with each candidate financial product. The filtering unit is used to filter out a preset number of candidate financial products from the financial product recommendation list, obtain the target financial product, and send the target financial product to the client used by the financial user.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored executable program, wherein, when the executable program is executed, it controls the device on which the computer-readable storage medium is located to perform the screening method for financial products according to any one of claims 1 to 7.

10. An electronic device, characterized in that, include: Memory, which stores executable programs; A processor for running the program, wherein the program, when running, performs the screening method for financial products according to any one of claims 1 to 7.

11. A computer program product comprising computer instructions, characterized in that, When the computer instructions are executed by the processor, they implement the steps of the screening method for financial products according to any one of claims 1 to 7.

Images