Personalized recommendation method and device, equipment, storage medium and program product
By combining deep learning models with user behavior and supply chain data, and dynamically updating the frequency, the problem of time-consuming user profile construction and inaccurate personalized recommendations caused by single-dimensional data is solved, realizing rapid response and accurate recommendations for user profiles.
Patent Information
- Application Number
- CN202511224728.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-08-29
- Publication Date
- 2025-12-16
AI Technical Summary
In existing technologies, user profile construction is time-consuming, lacks real-time performance, and relies on single-dimensional data, leading to inaccurate personalized recommendations. This is especially true for inclusive finance users such as SMEs, individual businesses, and low-activity users.
By constructing user profiles through deep learning models, combining user behavior data and supply chain data, dynamically updating the frequency, determining correlations and influence coefficients, and achieving personalized recommendations.
It shortens the time required to build user profiles, improves the accuracy and consistency of user profiles, enhances the precision of personalized recommendations, and adapts to changes in user needs.
Smart Images
Figure CN121144571A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the field of big data, and in particular to a personalized recommendation method and device, equipment, a storage medium and a program product. BACKGROUND
[0002] With the development of informationization process of the banking and financial industry, constructing a user portrait has become an important means to deeply mine user characteristics. With the help of a user portrait, the potential needs of users can be accurately analyzed, and personalized recommendation services can be provided to users, such as recommending goods or services to users.
[0003] At present, there are still some limitations in providing personalized recommendation services to users. On the one hand, a user portrait is usually constructed manually, which takes a long time to construct a user portrait, affects the real-time performance of a user portrait, and is difficult to reflect the potential needs of dynamic changes of users. At the same time, manually constructing a user portrait is highly subjective, and it is difficult to unify the standards for constructing a user portrait. On the other hand, it often relies on single-dimensional data, such as determining personalized recommendation services only according to transaction data. However, relying on single-dimensional data is usually not comprehensive in mining user needs, resulting in missing some needs of users in personalized recommendation. SUMMARY
[0004] The present application provides a personalized recommendation method, device, equipment, storage medium and program product, which uses a deep learning model to construct a user portrait, not only shortens the time for constructing a user portrait, but also unifies the standards for a user portrait. Based on the collected user behavior data and supply chain data, potential needs can be more accurately captured, and more accurate personalized recommendation can be achieved.
[0005] In a first aspect, the present application provides a personalized recommendation method, comprising:
[0006] Based on the historical user portrait of the target user, the portrait update frequency is determined, and the user data of the target user is collected according to the portrait update frequency, the user data at least including user behavior data and supply chain data; the association relationship of the target user is determined according to the supply chain data; the association relationship includes the associated user who trades with the target user, the user portrait of the associated user and the transaction type of the associated user and the target user; the user behavior data of the target user is input into the pre-trained portrait generation model to obtain the user portrait of the target user; the influence coefficient of the associated user is determined according to the transaction type of the associated user and the target user in the association relationship; the product or service is recommended to the target user according to the user portrait of the target user, the user portrait of the associated user and the influence coefficient of the associated user.
[0007] In a second aspect, the present application provides a personalized recommendation device, comprising:
[0008] The user data collection module is configured to determine a portrait update frequency based on a historical user portrait of a target user, and collect user data of the target user according to the portrait update frequency, wherein the user data at least includes user behavior data and supply chain data; the association relationship determination module is configured to determine an association relationship of the target user according to the supply chain data; the association relationship includes an associated user who trades with the target user, a user portrait of the associated user, and a transaction type of the associated user and the target user; the user portrait determination module is configured to input the user behavior data of the target user into a pre-trained portrait generation model to obtain a user portrait of the target user; the influence coefficient determination module is configured to determine an influence coefficient of the associated user according to the transaction type of the associated user and the target user in the association relationship; and the recommendation module is configured to recommend a product or a service to the target user according to the user portrait of the target user, the user portrait of the associated user, and the influence coefficient of the associated user.
[0009] In a third aspect, the present application provides an electronic device, comprising: a processor, and a memory connected with the processor in communication; the memory stores computer execution instructions; and the processor executes the computer execution instructions stored in the memory to implement the method provided in the first aspect.
[0010] In a fourth aspect, the present application provides a computer readable storage medium, wherein the computer readable storage medium stores computer execution instructions, and the computer execution instructions are executed by a processor to implement the method provided in the first aspect.
[0011] In a fifth aspect, the present application provides a computer program product, comprising a computer program, wherein the computer program is executed by a processor to implement the method provided in the first aspect.
[0012] The personalized recommendation method, device, equipment, storage medium and program product provided by the present application first determine the portrait update frequency according to the historical user portrait, collect user data according to the portrait update frequency during data collection, realize intelligent adjustment of the update frequency of the user portrait, and realize efficient use of resources while ensuring the timeliness of the user portrait; input the collected user behavior data into a portrait generation model to obtain a user portrait, which overcomes the problem of manually constructing a user portrait, and the user portrait obtained is more accurate due to the powerful learning and reasoning ability of the model; meanwhile, supply chain data is added to the user data, the association user of the user is determined by analyzing the supply chain data, and the recommendation is made based on the user portrait and the association user portrait and the influence coefficient of the association user during personalized recommendation, the data of the relationship network dimension of the user is supplemented by combining the supply chain data, the user is recommended from the perspective of the user who has an association with the user, the problem of incomplete mining of the user demand caused by relying on single-dimensional data is overcome, and the personalized recommendation is more accurate. BRIEF DESCRIPTION OF DRAWINGS
[0013] The accompanying drawings, which are incorporated herein and constitute part of the specification, illustrate embodiments consistent with the present application and, together with the description, further serve to explain the principles of the application.
[0014] Figure 1 A flowchart of a personalized recommendation method provided for an embodiment of the present application;
[0015] Figure 2 A flowchart of another personalized recommendation method provided for an embodiment of the present application;
[0016] Figure 3 A schematic diagram of an association relationship network provided for an embodiment of the present application;
[0017] Figure 4 A structural schematic diagram of a personalized recommendation device provided for an embodiment of the present application;
[0018] Figure 5 A structural schematic diagram of an electronic device provided for an embodiment of the present application.
[0019] The specific embodiments of the present application have been shown through the above-described drawings, and will be described in more detail hereinafter. These drawings and the written description are not intended to restrict the scope of the present application concept in any way, but to illustrate the present application concept to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION
[0020] The exemplary embodiments will be described in detail herein with reference to the attached drawings. The same numbers are used in different drawings to represent the same or similar elements. The following detailed description is not intended to restrict the embodiments to the embodiments described herein, but rather, is intended to explain the embodiments to those skilled in the art. Other embodiments can be utilized, and structural, logical and electrical changes can be made without departing from the scope of the present application.
[0021] It should be noted that the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of related data comply with relevant laws, regulations and standards of relevant countries and regions, necessary security measures are taken, do not violate public order and good customs, and provide corresponding operation portal for user to choose authorization or refusal.
[0022] And the present application relates to the big data analysis of user information (including but not limited to personal biological characteristics, identity data, consumption data, asset data, electronic terminal operation data, etc.), and the automatic decision-making by using artificial intelligence technology, and the technical scheme of making a decision with a significant impact on personal rights and interests based on the automatic decision-making result, providing the corresponding operation portal for the user to select to agree or reject the automatic decision-making result; if the user chooses to reject, the expert decision-making process is entered.
[0023] It should be noted that the personalized recommendation method, device, equipment, storage medium and program product provided by the present application can be used in the field of big data, and can also be used in any field other than the field of big data. The application field of the personalized recommendation method, device, equipment, storage medium and program product provided in the present application is not limited.
[0024] The specific application scenario of the present application is the scenario of customizing personalized recommendation for users by the bank financial system according to user data. Specifically, the bank financial system analyzes the data of the user, establishes a user portrait, and the user portrait is obtained by refining the real characteristics, behaviors, needs, etc. of the user in a systematic way, and the collection of user features; according to the user portrait, the potential needs of the user can be analyzed, and then according to the potential needs of the user, the products or services that meet the potential needs are recommended to the user.
[0025] For example, the behavior data of the user, such as transaction data, is obtained; according to the behavior data of the user, the purchase behavior, consumption habit and other characteristics of the user can be obtained, and the obtained characteristics can constitute the user portrait. According to the user portrait, products or services that meet the purchase preference, consumption frequency and other conditions of the user are recommended to the user. For example, products similar to the purchase preference are recommended to the user; if the user's consumption frequency is high, services that improve the convenience of consumption, such as automatic deduction service, can be recommended to the user.
[0026] For the application scenario of the present application, the prior art usually constructs a user portrait by artificial means. Since artificial data collection and analysis and integration of user data are time-consuming, the real-time performance and accuracy of the user portrait are affected. For example, if the user portrait is constructed for a long time, the characteristics of the user may have changed, resulting in an inaccurate user portrait. In addition, when the user portrait is constructed by artificial means, it depends on artificial experience, and the standards for constructing the user portrait by different workers are different. The inconsistency of the standards for constructing the user portrait of different users will affect the consistency of the user portrait of different users, and thus the personalized recommendation is inaccurate.
[0027] On the other hand, the prior art usually only uses single-dimensional user data, which may miss part of the characteristics of the user, and thus cannot comprehensively evaluate the potential needs of the user. For example, only using transaction data, the basic attributes of the user, such as the occupation of the user and the region where the user is located, may be missed, and thus the consumption differences between different occupations and different regions are ignored. Especially for users of inclusive finance, such as small and medium-sized and micro enterprises, individual businesses and low-activity users, which are insufficiently covered by financial services, the activity of such users is relatively low, and the user data is less. Using only single-dimensional user data, the user characteristics are not deeply mined, and thus the personalized recommendation finally provided is not accurate.
[0028] The personalized recommendation method provided by the present application aims to solve the above technical problems. Specifically, according to the determined portrait update frequency, the user data of the user is collected, and the user data at least includes user behavior data and supply chain data. By using multi-dimensional data, the user characteristics are not deeply mined; according to the supply chain data, the association relationship of the user is determined; according to the user behavior data, the user portrait of the user is obtained by using a deep learning model. Through the powerful ability of the model, not only the time for obtaining the user portrait is shortened, but also the user portrait is more unified; then, according to the association relationship, the influence coefficient of the associated user on the user is determined; finally, the user portrait of the user is combined with the user portrait of the associated user and the influence coefficient of the associated user, and the product or service is recommended to the user, so that more accurate personalized recommendation is realized; at the same time, according to the historical user portrait, the portrait update frequency is determined, and the user portrait is dynamically generated, so that the user demand change can be quickly responded.
[0029] The technical solutions of the present application and how the technical solutions of the present application solve the above technical problems will be described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of the present application will be described below with reference to the accompanying drawings.
[0030] Figure 1 A flowchart of a personalized recommendation method provided by an embodiment of the present application is shown. The personalized recommendation method provided by the present application can be executed by an electronic device with corresponding processing capability. As shown in the figure, the personalized recommendation method provided by the present application includes the following steps: Figure 1
[0031] Step S101, based on the historical user portrait of the target user, determine the portrait update frequency, and collect the user data of the target user according to the portrait update frequency.
[0032] The target user is a user who needs personalized recommendation, which can be an individual or an enterprise. In some embodiments, the target user is preferably a user of inclusive finance, such as a small and medium-sized micro enterprise, an individual industrial and commercial household, and a low-activity user. The user of inclusive finance has less interaction with the financial banking system, resulting in less user data. The historical user portrait is the current user portrait of the target user.
[0033] The user portrait is a dynamic view of at least one user feature generated by data fusion and analysis. The user feature can include an activity score, a behavior feature, and a risk score of the user.
[0034] The user data includes at least user behavior data and supply chain data. The user behavior data is a set of information generated by the target user when operating a digital product or service, and at least includes transaction information generated by the target user in the banking financial system. The supply chain data is used to reflect the transaction relationship between the target user and other users, and can be obtained from the data stored in the banking financial system. For example, the target user pays to an associated user, and the associated user pays to the target user. Through the supply chain data, the dimension of the user data is supplemented from the perspective of the supply chain, and the characteristics of the target user can be more comprehensively captured.
[0035] Based on the time characteristics of the user features in the initial portrait of the target user, the portrait update frequency is determined. For example, according to the activity of the target user at the current time (the more the target user operates, the higher the activity), the higher the activity of the target user, the higher the portrait update frequency; the lower the activity of the target user, the lower the portrait update frequency.
[0036] For example, the initial portrait update frequency can be set in advance; if the activity of the target user in the initial portrait is high, such as greater than or equal to a first threshold, the initial portrait update frequency is increased to obtain the portrait update frequency; if the activity of the target user in the initial portrait is low, such as less than a second threshold, the initial portrait update frequency is decreased to obtain the portrait update frequency; if the activity is in an intermediate state, such as less than the first threshold and greater than or equal to the second threshold, the initial portrait frequency is determined as the portrait update frequency.
[0037] In some embodiments, the target user can not have a historical user portrait, and if not, a standard user portrait can be used as the historical user portrait. The standard user portrait is a user portrait that can represent the characteristics of most users obtained by integrating and analyzing a plurality of user portraits.
[0038] By determining the portrait update frequency, the user portrait can be dynamically updated. Compared with the user portrait updated at a fixed frequency, the user portrait provided in the embodiment is more time-effective and can better reflect the current characteristics of the target user. At the same time, the waste of data processing resources caused by frequent updating of the user portrait is also avoided. For example, for users of inclusive finance, due to their low activity level, i.e., less transaction behavior, if the user portrait is updated at a higher frequency, it may lead to waste of data processing resources.
[0039] After determining the portrait update frequency, when the update time is reached, the subsequent steps are performed. Specifically, the user data of the target user is collected.
[0040] The execution subject of the method of the present application can be connected with the data source (such as the server of the bank financial system) to collect the user data of the target user through the data source. For example, it can be connected with the internal bank financial system to collect the user behavior data and supply chain data of the target user according to the operation records of each user stored by the bank financial system. It should be noted that the connection of the data source is performed on the premise that the target user and each party (such as the management party of the data source) have been fully authorized.
[0041] In some embodiments, since the data source can come from different servers, the data provided by different servers can be multi-source heterogeneous data, i.e., the data comes from different channels and the modalities of the data are different, which can be structured data, semi-structured data or unstructured data, etc. For example, the transaction information generated in the bank financial system is structured data or unstructured data, and the supply chain data can be a kind of image data with connection relationship, such as transaction relationship directed graph. In addition, part of the user data also contains time sequence relationship, such as transaction information. If there is multi-source heterogeneous data, the computer cannot directly analyze and process it.
[0042] In the present embodiment, a multi-modal data fusion method is proposed. Specifically, for unstructured data such as text data, natural language processing techniques are used, such as using a pre-trained language model to extract semantic features of the text data; for unstructured data such as image data, an OCR (Optical Character Recognition) information recognition is realized through a pre-trained feature extraction model; for data containing time sequence relationship, the data is discretized into statistical features of time window, such as weekly average transaction amount. Through multi-modal data fusion, it is more convenient for the computer to understand and process multi-source heterogeneous data.
[0043] In the present embodiment, since most of the user data of the target user collected is private data, in order to protect the security of the user data, after collecting the user data of the target user, the user data can be encrypted to prevent leakage of the user data.
[0044] For example, after collecting the user data of the target user, the user data is encrypted by using a privacy protection framework such as a federated learning framework, and the encrypted user data is uploaded to a related server or cloud to perform subsequent steps.
[0045] In step S102, the association relationship of the target user is determined according to the supply chain data.
[0046] The association relationship includes an associated user who trades with the target user, a user portrait of the associated user, and a transaction type of the associated user and the target user. The transaction type of the associated user and the target user is determined according to the transaction direction between the associated user and the target user in the supply chain data, such as whether the target user is a payer or a payee.
[0047] Specifically, according to the supply chain data, the user who has traded with the target user in the supply chain data is determined as the associated user, and the user portrait of the associated user is obtained; at the same time, the transaction type is determined according to the transaction direction between the associated user and the target user.
[0048] In step S103, the user behavior data of the target user is input into a pre-trained portrait generation model to obtain the user portrait of the target user.
[0049] The portrait generation model is a deep learning model. The portrait generation model is obtained by training the deep learning model. The training method of the deep learning model is to use the existing user portrait and the corresponding user behavior data as the training set, and the deep learning model learns the training set to finally obtain the trained portrait generation model. The specific training method can refer to the existing deep learning model training method, which will not be described here.
[0050] The user behavior data of the target user is input into the pre-trained portrait generation model, and the pre-trained portrait generation model analyzes and mines the user behavior data to output the user portrait of the target user.
[0051] In some embodiments, in order to improve the accuracy of the portrait generation model, the portrait generation model can be updated and optimized. Specifically, by using the newly generated user portrait, the parameters of the portrait generation model are adjusted by the incremental learning method.
[0052] In some embodiments, in order to accurately depict the dynamic changes of user features in the time dimension and improve the guarantee ability of the user portrait to the time sequence information, a real-time stream processing technology can be introduced into the portrait generation model. Specifically, the user behavior data is processed based on the real-time stream processing technology, and the processed user behavior data is input into the pre-trained user portrait determination module to obtain the user portrait including the time relationship.
[0053] In this embodiment, the training set for training the portrait generation model is the existing user portrait including time relationship and the user behavior data after real-time stream processing.
[0054] In step S104, the influence coefficient of the associated user is determined according to the transaction type of the associated user and the target user in the association relationship.
[0055] The corresponding relationship between the transaction type and the influence coefficient can be set in advance. For example, if the target user is the payer, the influence coefficient is the first coefficient; if the target user is the payee, the influence coefficient is the second coefficient.
[0056] According to the transaction type, the influence coefficient corresponding to the transaction type can be determined from the corresponding relationship between the transaction type and the influence coefficient.
[0057] In step S105, the product or service is recommended to the target user according to the user portrait of the target user, the user portrait of the associated user and the influence coefficient of the associated user.
[0058] The product or service can be each business or each service in the bank financial system, such as financial products, loan business, consumption installment service, etc.
[0059] Specifically, the user portrait of the associated user and the influence coefficient of each associated user are weighted, the weighted result is combined with the user portrait of the target user, matched with each product or service, and the matched product or service is recommended to the target user.
[0060] In an example, the weighted result and the user portrait of the target user can be input into a pre-trained product or service determination model to obtain a target product or a target service; the target product or the target service is recommended to the target user. The product or service determination model can be a deep learning model.
[0061] In another example, the information of each inventory user and the product or service successfully recommended to each inventory user can be obtained. The result of combining the weighted result with the user portrait of the target user is matched with the information of each inventory user, and the product or service corresponding to the inventory user satisfying the preset matching condition is recommended to the target user.
[0062] For example, the user portrait is combined with the user portrait of the associated user and the influence coefficient of the associated user, the user portrait is adjusted to obtain a user portrait combined with the association relationship. The user portrait combined with the association relationship is matched with the user portrait of the inventory user, and the product or service successfully recommended to the matched inventory user is recommended to the target user. The user portrait of the inventory user can be a stored user portrait or a user portrait combined with the association relationship obtained by the method provided in this embodiment.
[0063] In yet another example, a product or service can be established in correspondence with the user portrait of the associated user and the weighted result of the influence coefficient, and the corresponding relationship of the user portrait. According to the corresponding relationship, the matched product or service is determined and recommended to the target user.
[0064] The personalized recommendation method provided by the embodiment first determines the portrait update frequency according to the historical user portrait, collects user data according to the portrait update frequency during data collection, realizes intelligent adjustment of the update frequency of the user portrait, and realizes efficient use of resources while ensuring the timeliness of the user portrait; input the collected user behavior data into the portrait generation model to obtain the user portrait, which overcomes the problem of manually constructing the user portrait, and the user portrait obtained by the model is more accurate due to the powerful learning and reasoning ability of the model; meanwhile, the supply chain data is added to the user data, the associated user of the user is determined by analyzing the supply chain data, and the personalized recommendation is performed based on the user portrait, the influence coefficient of the associated user portrait and the associated user. By combining the supply chain data, the relationship network dimension of the user is completed, the user is recommended from the perspective of the associated user, the problem of incomplete user demand mining caused by relying on single-dimensional data is overcome, and the personalized recommendation is more accurate.
[0065] Figure 2 The flowchart of another personalized recommendation method provided by the embodiment is shown. The personalized recommendation method provided by the embodiment is based on Figure 1 The embodiment provides a personalized recommendation method, which is based on the embodiment shown in Figure 2 The personalized recommendation method provided by the embodiment includes the following steps:
[0066] Step S201, based on the historical user portrait of the target user, determine the portrait update frequency, and collect the user data of the target user according to the portrait update frequency.
[0067] The user data at least includes user behavior data and supply chain data.
[0068] Step S202, according to the supply chain data, construct an association relationship network with the target user as the center node.
[0069] The association relationship network is a network for representing the transaction behavior of the target user and the associated user, and is often represented in the form of a graph. The association relationship network includes multiple nodes and connection edges between the nodes. In addition to the center node representing the target user, each node in the association relationship network is an associated user. When there is a connection edge between two nodes, it means that there is a transaction behavior between the users represented by the two nodes. The connection edge can be a directed edge, indicating the direction of the fund flow.
[0070] For example, if there is supply chain data for the target user to pay the associated user 1, a directed connection edge from the target user node to the associated user 1 node is generated in the association relationship network.
[0071] Figure 3 An association relationship network diagram provided by an embodiment of the present application is shown in the figure. Figure 3 As shown, the association relationship network includes a target user G and associated users A1-A5. The associated users A1-A5 have direct or indirect transaction relationships with the target user G through connection edges. Specifically, the target user G has paid the associated users A1, A2 and A3; the associated users A4 and A3 have paid the target user G; and the associated user A5 has paid the associated user A4, thereby indirectly having a connection relationship with the target user G.
[0072] In the association relationship network, the properties of the connection edges can be adjusted according to the transaction frequency and transaction amount of the target user and the associated users. For example, connection edges of different colors or different thicknesses can be used to represent different ranges of transaction frequency and transaction amount.
[0073] In step S203, the association relationship and the association degree of each associated user and the target user are determined according to the association relationship network.
[0074] The association degree of the associated user and the target user is an index for measuring the transaction depth of the associated user and the target user, such as transaction frequency and transaction amount.
[0075] According to the association relationship network, the users corresponding to the nodes other than the target user in the association relationship network are determined as the associated users, and the user portraits of the associated users are obtained, such as the associated users A1-A5 in the figure. Figure 3 Then, the transaction type is determined through the direction of the connection edge between the nodes, and the association degree of the associated user and the target user is determined through the properties of the connection edge between the nodes, such as thickness and color.
[0076] In step S204, the user behavior data of the target user is input into a pre-trained portrait generation model to obtain the user portrait of the target user.
[0077] In step S205, the association degree of the associated user and the target user is obtained.
[0078] In step S206, the influence coefficient of the associated user is determined based on the transaction type and the association degree of the associated user and the target user.
[0079] Specifically, the transaction type and the association degree are weighted to obtain the influence coefficient of the associated user.
[0080] For example, the influence coefficient R can be expressed as R = λ1 T + λ2 D.
[0081] wherein T is a transaction type, the transaction type can be set as a target user being a payer, and the transaction type is assigned a first value; the transaction type can be set as the target user being a payee, and the transaction type is assigned a second value; D is a correlation degree; λ1 and λ2 are weight values corresponding to the transaction type and the correlation degree respectively, and are configurable parameters.
[0082] Optionally, the transaction type includes an upstream transaction and a downstream transaction, the upstream transaction is a transaction in which the target user is a payer, and the downstream transaction is a transaction in which the target user is a payee; based on the transaction type and the correlation degree between the associated user and the target user, the influence coefficient of the associated user is determined, including: obtaining an initial influence coefficient; if the transaction type between the associated user and the target user is the upstream transaction, then the initial influence coefficient is reduced according to the correlation degree of the associated user to obtain the influence coefficient of the associated user; if the transaction type between the associated user and the target user is the downstream transaction, then the initial influence coefficient is increased according to the correlation degree of the associated user to obtain the influence coefficient of the associated user.
[0083] In this embodiment, the initial influence coefficient can also be set in advance. The initial influence coefficient is a configurable parameter set in advance. The data of a plurality of users can be counted, the influence degree of the associated user on the user is analyzed, and a parameter value representing a majority of the population is set according to the influence degree, such as an average value of the influence degree.
[0084] Compared with the upstream transaction and the downstream transaction, the influence degree of the associated user of different transaction types on the target user is different. For example, if the associated user of the downstream transaction has credit exceptions, there may be a phenomenon of arrears of payment, which seriously affects the circulation of funds of the target user.
[0085] In this application, the initial influence coefficient is adjusted according to the associated user of different transaction types. Specifically, if it is an upstream transaction, the initial influence coefficient is reduced, and the reduction amplitude of the initial influence coefficient is determined according to the correlation degree, such as the higher the correlation degree, the greater the reduction amplitude, and finally the influence coefficient of the associated user is obtained. If it is a downstream transaction, the initial influence coefficient is increased, and the increase amplitude of the initial influence coefficient is determined according to the correlation degree, such as the higher the correlation degree, the greater the increase amplitude, and finally the influence coefficient of the associated user is obtained.
[0086] In actual application, the amplitude of the initial influence coefficient determined according to the correlation degree of the reduction or increase can be determined according to actual conditions, which is not limited herein.
[0087] The method is provided with a clear benchmark value by setting the initial influence coefficient, so that extreme situations caused by sudden changes of some parameters are avoided; and the initial influence coefficient is adjusted according to actual association conditions, so that the influence coefficient is more accurate on the basis of the benchmark value.
[0088] In step S207, the user portraits of the target user and the associated users are fused based on the influence coefficients of the associated users to obtain a fused portrait.
[0089] Specifically, the influence coefficients of the associated users are weighted and averaged with the user portraits of the associated users to obtain an associated portrait of the target user; and then the associated portrait is fused with the user portrait of the target user, such as weighted, to obtain the fused portrait. The weights of the portraits are pre-set configurable parameters.
[0090] For example, the user portrait includes an activity score, a risk score and a transaction intensity. First, the activity score, the risk score and the transaction intensity of each associated user are weighted and averaged with the influence coefficient to obtain a first activity score, a first risk score and a first transaction intensity of the associated portrait; then, the first activity score is weighted with the activity score of the user portrait to obtain an activity score of the fused portrait; the first risk score is weighted with the activity score of the user portrait to obtain an activity score of the fused portrait; and the first transaction intensity score is weighted with the transaction intensity score of the user portrait to obtain an activity score of the fused portrait.
[0091] Optionally, the user portrait includes an activity score, a risk score and a transaction intensity; and the risk score in the fused portrait is a weighted result of the risk scores of the user portraits of the target user and the associated users, and the weight in the weighting is determined by the influence coefficient.
[0092] The influence of the associated user on the target user can be only for part of the features, such as the risk score, while the influence of the associated user on the target user is less for the transaction intensity and the activity score. In this embodiment, the fusion can be performed only for part of the features.
[0093] In this embodiment, only the risk score is fused. Specifically, the risk score in the fused portrait can be directly obtained by weighting the risk score of the target user with the risk score of each associated user.
[0094] The weight in the weighting can be determined according to the influence coefficient. For example, the weight of the risk score of the user portrait of each associated user is the product of the influence coefficient and a first parameter; the weight of the risk score of the user portrait of the target user is a second parameter; and the first parameter and the second parameter are pre-set configurable parameters.
[0095] In step S208, a product or a service is recommended to the target user based on the fused portrait.
[0096] Specifically, the fusion portrait is matched with each product or service, and the matched product or service is recommended to the target user.
[0097] Specifically, the fusion portrait is matched with each product or service, and the matched product or service is recommended to the target user.
[0098] Optionally, the method provided by the embodiment further includes: if the risk score of the user portrait of the target user meets a preset risk condition, or the risk score of the fusion portrait meets a preset fusion risk condition, a risk warning of the target user is performed.
[0099] In the embodiment, in addition to the recommendation to the target user, a risk warning can also be performed to ensure the safety of the banking and financial system.
[0100] Specifically, if it is detected that the risk score of the user portrait of the target user meets a preset risk condition, such as the risk score being greater than or equal to a first risk value, a risk warning of the target user is performed; or, if it is detected that the risk score of the fusion portrait of the target user meets a preset fusion risk condition, such as the risk score being greater than or equal to a second risk value, a risk warning of the target user is performed.
[0101] Step S209: adjusting the portrait update frequency according to the user portrait of the target user.
[0102] Specifically, according to the activity score in the user portrait of the target user, if the activity score is high at the current time, the portrait update frequency can be increased; if the activity score is low at the current time, the portrait update frequency can be decreased.
[0103] Optionally, adjusting the portrait update frequency according to the user portrait of the target user includes: determining an activity trend feature of the target user according to the user portrait of the target user and the historical user portrait; and adjusting the portrait update frequency according to the activity trend feature of the target user.
[0104] The activity trend feature is the change of the activity of the target user in the time dimension. For example, the activity of the target user is gradually reduced or increased.
[0105] Specifically, the activity trend is determined according to the comparison result of the activity index of the user portrait of the target user and the activity index of the historical user portrait, and the activity trend feature of the target user is determined. If the trend feature is reduced, the portrait update frequency is increased; if the trend feature is increased, the portrait update frequency is decreased.
[0106] By determining the activity trend characteristics, adjusting the portrait update frequency, the transaction behavior change of the target user can be predicted in advance, so that the portrait update frequency is more reasonable, and resource waste caused by frequent updating is avoided, and the immediate demand of the target user is missed due to long-time non-updating.
[0107] By constructing the association relationship transaction network, the association relationship and the association degree of the target user can be obtained more intuitively and deeply, and the complexity of obtaining the association relationship is reduced; and by determining the influence coefficient of the associated user on the target user according to the association degree, the influence coefficient not only represents the transaction type of the associated user and the target user, but also accurately represents the association depth of the associated user and the target user; at the same time, by obtaining the fused portrait, the user feature in the user behavior data and the supply chain data is combined, and the product or service recommended to the user can be determined through the information of one portrait, so as to avoid data redundancy and conflict between two user portraits, and enhance the data analysis capability; in addition, according to the newly obtained user portrait, the portrait update frequency is dynamically adjusted, so that the user portrait can reflect the latest user behavior and preference, thereby improving the accuracy of personalized recommendation.
[0108] Figure 4 A structural schematic diagram of a personalized recommendation device provided by an embodiment of the present application is shown in FIG. 1. Figure 4 As shown in FIG. 1, the personalized recommendation device provided by the embodiment includes a user data acquisition module 401, an association relationship determination module 402, a user portrait determination module 403, an influence coefficient determination module 404, and a recommendation module 405.
[0109] The user data acquisition module 401 is configured to determine the portrait update frequency based on the historical user portrait of the target user, and acquire the user data of the target user according to the portrait update frequency, wherein the user data at least includes user behavior data and supply chain data; the association relationship determination module 402 is configured to determine the association relationship of the target user according to the supply chain data; the association relationship includes the associated user who trades with the target user, the user portrait of the associated user, and the transaction type of the associated user and the target user; the user portrait determination module 403 is configured to input the user behavior data of the target user into a pre-trained portrait generation model to obtain the user portrait of the target user; the influence coefficient determination module 404 is configured to determine the influence coefficient of the associated user according to the transaction type of the associated user and the target user in the association relationship; and the recommendation module 405 is configured to recommend products or services to the target user according to the user portrait of the target user, the user portrait of the associated user, and the influence coefficient of the associated user.
[0110] Optionally, the recommendation module 405 is specifically configured to:
[0111] The user portrait of the target user and the associated user is fused based on the influence coefficient of the associated user to obtain a fused portrait; and the product or service is recommended to the target user based on the fused portrait.
[0112] Optionally, the user portrait includes an activity score, a risk score and a transaction intensity; and the risk score in the fused portrait is a weighted result of the risk scores of the user portraits of the target user and the associated user, and the weight in the weighting is determined by the influence coefficient.
[0113] Optionally, the influence coefficient determination module 404 is specifically configured to:
[0114] obtain the association degree between the associated user and the target user; and determine the influence coefficient of the associated user based on the transaction type between the associated user and the target user and the association degree.
[0115] Optionally, the transaction type includes an upstream transaction and a downstream transaction, the upstream transaction is that the target user is a payer, and the downstream transaction is that the target user is a payee; and the influence coefficient determination module 404 is specifically configured to:
[0116] obtain the association degree between the associated user and the target user; obtain an initial influence coefficient; if the transaction type between the associated user and the target user is the upstream transaction, reduce the initial influence coefficient according to the association degree of the associated user to obtain the influence coefficient of the associated user; and if the transaction type between the associated user and the target user is the downstream transaction, increase the initial influence coefficient according to the association degree of the associated user to obtain the influence coefficient of the associated user.
[0117] Optionally, the association relationship determination module 402 is specifically configured to:
[0118] construct an association relationship network with the target user as a center node based on the supply chain data, each node in the association relationship network being an associated user; and determine the association relationship and the association degree between each associated user and the target user based on the association relationship network.
[0119] Optionally, the personalized recommendation device further includes a portrait update frequency adjustment module, which is configured to adjust the portrait update frequency according to the user portrait of the target user.
[0120] Optionally, the portrait update frequency adjustment module is specifically configured to:
[0121] determine an activity trend feature of the target user based on the user portrait of the target user and a historical user portrait; and adjust the portrait update frequency according to the activity trend feature of the target user.
[0122] The personalized recommendation device provided by the embodiments of the present application can be used to execute the technical solutions of the personalized recommendation method provided by any of the embodiments of the present application, and the implementation principles and technical effects are similar, and the present embodiment will not be repeated here.
[0123] Figure 5 A structural schematic diagram of an electronic device is provided in the embodiments of the present application. As shown in the figure, Figure 5 The electronic device of the present embodiment can include at least one processor 501, and a memory 502 in communication connection with the at least one processor; wherein the memory 502 stores instructions executable by the at least one processor 501, and the instructions are executed by the at least one processor 501 to enable the electronic device to execute the method of any of the above embodiments.
[0124] Optionally, the memory 502 can be independent or integrated with the processor 501. When the memory 702 is independently provided, the device further includes a bus for connecting the memory 502 and the processor 501.
[0125] The implementation principles and technical effects of the electronic device provided by the present embodiment can be referred to the foregoing embodiments, and will not be repeated here.
[0126] The embodiments of the present application further provide a computer readable storage medium, and the computer readable storage medium stores computer execution instructions, and when the computer execution instructions are executed by a processor, the method provided by any of the foregoing embodiments can be implemented.
[0127] The embodiments of the present application further provide a computer program product, which includes a computer program, and the computer program is executed by a processor to implement the method provided by any of the foregoing embodiments.
[0128] It should be noted that, for the foregoing method embodiments, in order to simply describe, they are all described as a series of action combinations, but those skilled in the art should know that the present application is not limited to the action sequence described, because according to the present application, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should know that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily required by the present application.
[0129] It should be further noted that although the steps in the flowchart are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowchart may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.
[0130] It should be understood that the above-described device embodiments are merely illustrative, and the device of this application can also be implemented in other ways. For example, the division of units / modules in the above embodiments is only a logical functional division, and there may be other division methods in actual implementation. For example, multiple units, modules, or components may be combined, or integrated into another system, or some features may be ignored or not executed.
[0131] Furthermore, unless otherwise specified, the functional units / modules in the various embodiments of this application can be integrated into one unit / module, or each unit / module can exist physically separately, or two or more units / modules can be integrated together. The integrated units / modules described above can be implemented in hardware or as software program modules.
[0132] When integrated units / modules are implemented in hardware, the hardware can be digital circuits, analog circuits, etc. The physical implementation of the hardware structure includes, but is not limited to, transistors, memristors, etc. Unless otherwise specified, the processor can be any suitable hardware processor, such as a CPU, GPU, FPGA, DSP, and ASIC, etc. Unless otherwise specified, the storage unit can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc.
[0133] If the integrated units / modules are implemented in the form of software program modules and sold or used as independent products, they can be stored in a computer readable memory. Based on such understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a memory and includes a number of instructions for causing a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the embodiments of the method of the present application. The aforementioned memory includes: a U disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a mobile hard disk, a magnetic disk or an optical disk, and various media that can store program codes.
[0134] In the above embodiments, the description of each embodiment has its own focus, and the parts not described in detail in a certain embodiment can be referred to the related description of other embodiments. The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above embodiments are described, however, as long as the combinations of the technical features do not exist contradictory, they should be considered as the scope of the present application.
[0135] Other embodiments of the application will be apparent to those skilled in the art from consideration of the specification and practice of the application disclosed herein. The application is intended to cover any variations, uses or adaptations of the application following, in general, the principles of the application and including such departures from the present disclosure as come within known or customary practice in the art to which the application pertains or can relate. The specification and examples are to be regarded as exemplary only, and the true scope and spirit of the application are indicated by the following claims.
[0136] It should be understood that the application is not limited to the precise construction that has been described above and illustrated in the accompanying drawings, and that various modifications and changes can be made without departing from the scope thereof. The scope of the application is limited only by the claims that follow.
Claims
1. A personalized recommendation method, characterized by, The method comprises: determining an image update frequency based on a historical user image of a target user, and collecting user data of the target user according to the image update frequency, wherein the user data at least includes user behavior data and supply chain data; determining an association relationship of the target user according to the supply chain data, wherein the association relationship includes an associated user who trades with the target user, a user image of the associated user, and a transaction type between the associated user and the target user; inputting the user behavior data of the target user into a pre-trained image generation model to obtain a user image of the target user; determining an influence coefficient of the associated user according to the transaction type between the associated user and the target user in the association relationship; recommending a product or a service to the target user according to the user image of the target user, the user image of the associated user, and the influence coefficient of the associated user.
2. The method of claim 1, wherein, The step of recommending a product or a service to the target user according to the user image of the target user, the user image of the associated user, and the influence coefficient of the associated user comprises: fusing the user images of the target user and the associated user based on the influence coefficient of the associated user to obtain a fused image; recommending a product or a service to the target user based on the fused image.
3. The method of claim 2, wherein, The user image includes an activity score, a risk score, and a transaction intensity; the risk score in the fused image is a weighted result of the risk scores of the user images of the target user and the associated user, and the weight in the weighting is determined by the influence coefficient.
4. The method of claim 3, wherein, The method further comprises: if the risk score of the user image of the target user meets a preset risk condition, or the risk score of the fused image meets a preset fused risk condition, performing a risk warning for the target user.
5. The method of claim 1, wherein, The step of determining the influence coefficient of the associated user according to the transaction type between the associated user and the target user in the association relationship comprises: obtaining an association degree between the associated user and the target user; determining the influence coefficient of the associated user based on the transaction type between the associated user and the target user and the association degree.
6. The method of claim 5, wherein, The transaction type includes an upstream transaction and a downstream transaction, the upstream transaction is that the target user is a payer, and the downstream transaction is that the target user is a payee; the step of determining the influence coefficient of the associated user based on the transaction type between the associated user and the target user and the association degree comprises: obtaining an initial influence coefficient; if the transaction type between the associated user and the target user is the upstream transaction, reducing the initial influence coefficient according to the association degree of the associated user to obtain the influence coefficient of the associated user; if the transaction type between the associated user and the target user is the downstream transaction, increasing the initial influence coefficient according to the association degree of the associated user to obtain the influence coefficient of the associated user.
7. The method of claim 5, wherein, The step of determining the association relationship of the target user according to the supply chain data comprises: constructing an association relationship network with the target user as a center node based on the supply chain data, wherein each node in the association relationship network is an associated user. According to the association relationship network, the association relationship and the association degree of each association user and the target user are determined.
8. The method according to any one of claims 1 to 7, characterized in that, The method further comprises: According to the user portrait of the target user, the portrait update frequency is adjusted.
9. The method of claim 8, wherein, According to the user portrait of the target user, the portrait update frequency is adjusted. According to the user portrait of the target user and the historical user portrait, the activity trend feature of the target user is determined; According to the activity trend feature of the target user, the portrait update frequency is adjusted.
10. A personalized recommendation device, characterized in that, Comprise: The user data acquisition module is used for determining the portrait update frequency based on the historical user portrait of the target user, and acquiring the user data of the target user according to the portrait update frequency, wherein the user data at least includes user behavior data and supply chain data; The association relationship determination module is used for determining the association relationship of the target user according to the supply chain data; the association relationship includes the association user who trades with the target user, the user portrait of the association user and the transaction type of the association user and the target user; The user portrait determination module is used for inputting the user behavior data of the target user into the pre-trained portrait generation model to obtain the user portrait of the target user; The influence coefficient determination module is used for determining the influence coefficient of the association user according to the transaction type of the association user and the target user in the association relationship; The recommendation module is used for recommending products or services to the target user according to the user portrait of the target user, the user portrait of the association user and the influence coefficient of the association user.
11. An electronic device, comprising: Comprise: A processor and a memory connected with the processor in communication; The memory stores computer execution instructions; The processor executes the computer execution instructions stored in the memory to realize the method of any one of claims 1 to 9.
12. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer execution instructions, and the computer execution instructions are executed by the processor to realize the method of any one of claims 1 to 9.
13. A computer program product, characterised in that, The computer program is executed by the processor to realize the method of any one of claims 1 to 9.