Personalized recommendation method and device based on user portrait

Abstract

The invention provides a personalized recommendation method based on a user portrait, and is applied to the field of artificial intelligence. The method comprises the following steps: acquiring multi-dimensional data of a user within a first preset time; wherein the multi-dimensional data comprises user basic information, behavior data, time data and a topological relation; performing feature extraction on the multi-dimensional data, and fusing the extracted features to obtain fused features; carrying out first dimension reduction processing on the fusion features, and carrying out second dimension reduction processing on the fusion features in a second preset time; wherein the first preset time is within second preset time, and the interval between the second preset time and the current time is smaller than or equal to a preset time threshold value; performing weighted fusion on the fusion features after the first dimension reduction processing and the second dimension reduction processing to obtain a user portrait; and recommending a corresponding product to the user based on the user portrait. The invention further provides a personalized recommendation device and equipment based on the user portrait, a storage medium and a program product.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence, specifically to the field of personalized product recommendation, and more specifically to a personalized recommendation method, apparatus, device, medium, and program product based on user portraits. Background Art

[0002] With the development of big data and artificial intelligence technologies, personalized recommendation methods based on customer profiles have emerged. However, with the continuous accumulation of customer data, the data dimension has increased dramatically, which has brought huge challenges to data analysis and customer profile construction. High-dimensional data not only increases the difficulty of calculation, but also makes the data analysis results difficult to interpret and apply. Existing customer profile construction algorithms include clustering algorithms and principal component analysis algorithms, but these algorithms have the following disadvantages: (1) they cannot effectively extract key information from high-dimensional customer data, thus affecting the accuracy and completeness of customer profiles; (2) the customer profiles generated by these two algorithms are difficult to correspond to actual business scenarios or customer characteristics and lack interpretability. Based on this, how to construct accurate customer profiles for high-dimensional customer data and how to provide personalized recommendations to customers based on customer profiles are currently urgent issues that need to be solved. Summary of the Invention

[0003] In view of the above problems, the present application provides a personalized recommendation method, device, equipment, medium and program product based on user portraits to improve the accuracy of user portrait construction and product recommendation precision.

[0004] According to a first aspect of the present application, a personalized recommendation method based on user portrait is provided, comprising: obtaining multi-dimensional data of a user within a first preset time; wherein the multi-dimensional data includes basic user information, behavior data, time data, and topological relationships;

[0005] Performing feature extraction on the multi-dimensional data, and fusing the extracted features to obtain fused features;

[0006] Performing a first dimensionality reduction process on the fused features, and performing a second dimensionality reduction process on the fused features within a second preset time; wherein the first preset time is within the second preset time and the interval between the second preset time and the current time is less than or equal to a preset time threshold;

[0007] Performing weighted fusion on the fused features after the first dimensionality reduction processing and the second dimensionality reduction processing to obtain a user profile;

[0008] Based on the user portrait, corresponding products are recommended to the user.

[0009] According to an embodiment of the present application, extracting features from the multi-dimensional data includes:

[0010] Normalizing the numerical data in the multi-dimensional data to obtain standardized numerical features;

[0011] Performing topic analysis on the text data in the multidimensional data using a linear discriminant analysis algorithm to extract topic distribution features of the text data;

[0012] Time series features are extracted from the data containing time series characteristics in the multi-dimensional data to obtain time series features.

[0013] According to an embodiment of the present application, the features extracted by fusion include:

[0014] Mapping the extracted features to the same dimension respectively;

[0015] The attention mechanism is used to perform weighted fusion of the features in the same dimension.

[0016] According to an embodiment of the present application, performing a first dimensionality reduction process on the fusion feature includes:

[0017] Using a uniform manifold approximation projection algorithm, constructing a high-dimensional graph according to the fused features;

[0018] Calculating the cross entropy between the high-dimensional graph and a randomly generated low-dimensional graph;

[0019] Optimizing the low-dimensional graph according to the cross entropy, and replacing the randomly generated low-dimensional graph with the optimized low-dimensional graph;

[0020] Repeat the cross entropy calculation and low-dimensional graph optimization operations until the obtained cross entropy is less than or equal to the preset cross entropy threshold;

[0021] The fused features are projected into a first low-dimensional subspace constructed by a low-dimensional graph corresponding to a cross entropy that is less than or equal to a preset cross entropy threshold.

[0022] According to an embodiment of the present application, performing a second dimensionality reduction process on the fusion features within the second preset time includes:

[0023] Calculating the covariance matrix of the fusion features within the second preset time by a principal component analysis algorithm;

[0024] Performing eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and corresponding eigenvectors;

[0025] Sorting the eigenvalues ​​in descending order, and constructing a principal component matrix based on the eigenvectors corresponding to the first preset number of eigenvalues ​​that are ranked first;

[0026] The fused features within the second preset time are projected into a second low-dimensional subspace formed by the principal component matrix.

[0027] According to an embodiment of the present application, recommending corresponding products to the user based on the user portrait includes:

[0028] Assessing the user's risk level based on the user profile;

[0029] If the user risk level is lower than a preset threshold, obtaining a first historical user profile that matches the user profile, and calculating the similarity between the user profile and the product profile corresponding to the first historical user profile;

[0030] If the similarity is greater than a preset similarity threshold, the product corresponding to the product portrait is recommended to the user.

[0031] According to an embodiment of the present application, the method further includes:

[0032] For new users, historically recommended products are sorted in descending order of recommendation frequency, and the second preset number of products ranked first are selected as the candidate product set;

[0033] Obtaining the second historical user portraits associated with each product in the candidate product set, respectively, calculating the similarity between the new user portrait and the second historical user portrait, and obtaining a third historical user that matches the new user;

[0034] Based on the recommended products received by the third historical user, a first recommended product is generated and recommended to the new user.

[0035] A second aspect of the present application provides a personalized recommendation device based on user portraits, comprising:

[0036] A multi-dimensional data acquisition module is used to acquire multi-dimensional data of the user within a first preset time; wherein the multi-dimensional data includes basic user information, behavior data, time data and topological relationships;

[0037] A fusion feature acquisition module is used to extract features from the multi-dimensional data and fuse the extracted features to obtain fusion features;

[0038] a dimensionality reduction processing module, configured to perform a first dimensionality reduction processing on the fused features, and perform a second dimensionality reduction processing on the fused features within a second preset time; wherein the first preset time is included in the second preset time and the interval between the second preset time and the current time is less than or equal to a preset time threshold;

[0039] A user portrait acquisition module, configured to perform weighted fusion on the fused features after the first dimensionality reduction processing and the second dimensionality reduction processing to obtain a user portrait;

[0040] The product recommendation module is used to recommend corresponding products to users based on the user portrait.

[0041] The third aspect of the present application provides an electronic device, comprising: one or more processors; a memory for storing one or more computer programs, wherein the one or more processors execute the one or more computer programs to implement the steps of the above method.

[0042] The fourth aspect of the present application further provides a computer-readable storage medium having a computer program or instructions stored thereon, which implements the steps of the above method when the computer program or instructions are executed by a processor.

[0043] The fifth aspect of the present application further provides a computer program product, comprising a computer program or instructions, which implement the steps of the above method when executed by a processor. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The above contents and other objects, features and advantages of the present application will become more apparent through the following description of the embodiments of the present application with reference to the accompanying drawings, in which:

[0045] Figure 1 A diagram schematically illustrates an application scenario of a personalized recommendation method, apparatus, device, medium, and program product based on user portraits according to an embodiment of the present application;

[0046] Figure 2 The following schematically shows a flow chart of a personalized recommendation method based on user portrait according to an embodiment of the present application;

[0047] Figure 3 The following schematically shows a first dimensionality reduction process flow chart according to an embodiment of the present application;

[0048] Figure 4 The following schematically shows a second dimensionality reduction process flow chart according to an embodiment of the present application;

[0049] Figure 5 The following schematically illustrates a product recommendation flow chart based on user portraits according to an embodiment of the present application;

[0050] Figure 6 A flowchart of product recommendation for new users according to an embodiment of the present application is schematically shown;

[0051] Figure 7 A block diagram schematically illustrates a structure of a personalized recommendation device based on user portraits according to an embodiment of the present application; and

[0052] Figure 8 A block diagram of an electronic device suitable for implementing a personalized recommendation method based on user portraits according to an embodiment of the present application is schematically shown. DETAILED DESCRIPTION

[0053] Hereinafter, embodiments of the present application will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present application. In the detailed description below, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present application. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessarily confusing the concepts of the present application.

[0054] The terms used herein are only for describing specific embodiments and are not intended to limit this application. The terms "comprise," "include," etc. used herein indicate the presence of the features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0055] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0056] When expressions such as "at least one of A, B, and C, etc." are used, they should generally be interpreted in accordance with the meaning commonly understood by those skilled in the art (for example, "a system having at least one of A, B, and C" should include but is not limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, C, etc.).

[0057] In the technical solution of this application, the user information involved (including but not limited to user personal information, user image information, user device information, such as location information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) are all 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 comply with relevant laws, regulations and standards, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0058] The embodiments of the present application provide a personalized recommendation method based on user portraits, which extracts features from multi-dimensional data to obtain fused features, performs dimensionality reduction processing in different time periods based on the fused features, generates user portraits based on the features after the two dimensionality reduction processing, and provides personalized recommendations to users based on the user portraits. This not only improves the accuracy of user portrait construction, but also improves risk control capabilities, thereby improving user satisfaction and the bank's service efficiency.

[0059] Figure 1 The application scenario diagram of the personalized recommendation method, apparatus, device, medium and program product based on user portrait according to an embodiment of the present application is schematically shown.

[0060] like Figure 1 As shown, the application scenario 100 according to this embodiment may include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104, and a server 105. The network 104 is used as a medium for providing a communication link between the first terminal device 101, the second terminal device 102, the third terminal device 103, and the server 105. The network 104 may include various connection types, such as wired or wireless communication links or optical fiber cables.

[0061] A user may use a first terminal device 101, a second terminal device 102, or a third terminal device 103 to interact with a server 105 via a network 104 to receive or send messages, etc. Various communication client applications may be installed on the first terminal device 101, the second terminal device 102, or the third terminal device 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (for example only).

[0062] The first terminal device 101 , the second terminal device 102 , and the third terminal device 103 may be various electronic devices having display screens and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, desktop computers, and the like.

[0063] The server 105 may be a server that provides various services, such as a background management server (for example only) that supports websites browsed by users using the first terminal device 101, the second terminal device 102, and the third terminal device 103. The background management server may analyze and process received data such as user requests, and feed back processing results (e.g., web pages, information, or data obtained or generated based on user requests) to the terminal devices.

[0064] It should be noted that the personalized recommendation method based on user portrait provided in the embodiment of the present application can generally be executed by the server 105. Accordingly, the personalized recommendation device based on user portrait provided in the embodiment of the present application can generally be set in the server 105. The personalized recommendation method based on user portrait provided in the embodiment of the present application can also be executed by a server or server cluster that is different from the server 105 and can communicate with the first terminal device 101, the second terminal device 102, the third terminal device 103 and / or the server 105. Accordingly, the personalized recommendation device based on user portrait provided in the embodiment of the present application can also be set in a server or server cluster that is different from the server 105 and can communicate with the first terminal device 101, the second terminal device 102, the third terminal device 103 and / or the server 105.

[0065] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is merely illustrative. Any number of terminal devices, networks and servers may be provided as required.

[0066] The following will be based on Figure 1 The scene described by Figures 2 to 6 A detailed description of the personalized recommendation method based on user portrait according to an embodiment of the present application is given.

[0067] Figure 2 The following schematically shows a flowchart of a personalized recommendation method based on user portrait according to an embodiment of the present application.

[0068] like Figure 2 As shown, the personalized recommendation method 200 based on user portrait of this embodiment includes operations S210 to S250.

[0069] In operation S210, multi-dimensional data of a user within a first preset time is obtained; wherein the multi-dimensional data includes basic user information, behavior data, time data, and topological relationships.

[0070] In operation S220 , feature extraction is performed on the multi-dimensional data, and the extracted features are fused to obtain fused features.

[0071] In operation S230, a first dimensionality reduction process is performed on the fused features, and a second dimensionality reduction process is performed on the fused features within a second preset time; wherein the first preset time is included in the second preset time and the interval between the second preset time and the current time is less than or equal to a preset time threshold.

[0072] In operation S240, weighted fusion is performed on the fused features after the first dimensionality reduction processing and the second dimensionality reduction processing to obtain a user portrait.

[0073] In operation S250, corresponding products are recommended to the user based on the user portrait.

[0074] In some embodiments, in operation S210, the first preset time may be a longer period such as 1 year, 2 years, or 3 years;

[0075] User basic information includes user name, age, date of birth, ID number, mobile phone number, bank account information and asset information, etc.;

[0076] Behavioral data includes data left by users through various operations or behaviors, such as transaction behavior data, product usage behavior data, and repayment behavior data. Transaction behavior data may include transaction amount, transaction frequency, and transaction type (such as transfer, withdrawal, or consumption). Product usage behavior data may include the frequency or duration of browsing financial products, the time or location of browsing financial products, etc. Repayment behavior data may include repayment date, repayment amount, and whether it is overdue.

[0077] Time data can include transaction time data, the date a user purchases financial products or insurance products, or the date a user's account is active.

[0078] Topological relationships include the associations between users and other entities (such as other users, devices, or accounts). These associations can be one or a combination of multiple relationships, such as subordinate relationships, cooperative relationships, and fund transaction relationships. For example, a company's public account has a fund transaction relationship with multiple personal accounts, a user's device has a binding relationship with the user account, and multiple users have a fund transaction relationship. If user A recommends user B to apply for a credit card, then a topological relationship is formed between user A and user B.

[0079] In some embodiments, in operation S230, the second preset time may be the time closest to the current time, for example, the past 7 days, the past 5 days, the past 3 days, etc. The preset time threshold may be determined according to actual conditions.

[0080] According to an embodiment of the present application, by performing dimensionality reduction processing on the fused features within the time period closest to the current time (i.e., within the second preset time), it is possible to determine whether the user has abnormal behavior recently, such as sudden large-amount transfers, transaction times suddenly concentrated in the early morning, etc., thereby providing personalized recommendations to customers while realizing risk control warnings, significantly improving the insight into customer status, and thus helping banks avoid risks; by performing dimensionality reduction processing on the fused features within a longer period of time (i.e., within the first preset time), it is possible to identify users' long-term preference patterns, etc., which is conducive to recommending long-term products to users, thereby enhancing user stickiness; fusing the fused features after these two dimensionality reduction processes can not only improve risk control capabilities, but also improve the accuracy of user portrait construction, thereby improving user satisfaction and bank service efficiency.

[0081] In some embodiments, feature extraction is performed on multi-dimensional data, including:

[0082] Standardize the numerical data in the multi-dimensional data to obtain standardized numerical features;

[0083] Use the linear discriminant analysis algorithm to perform topic analysis on text data in multi-dimensional data and extract the topic distribution characteristics of text data;

[0084] Time series features are extracted from multi-dimensional data containing time series characteristics to obtain time series features.

[0085] In some embodiments, before performing feature extraction on multi-dimensional data, the multi-dimensional data needs to be preprocessed. The preprocessing includes cleaning, removing noise, and removing outliers to improve the accuracy of the data.

[0086] In some embodiments, a linear discriminant analysis algorithm is used to perform topic analysis on text data in multi-dimensional data to extract topic distribution features of the text data, including:

[0087] The linear discriminant analysis algorithm is used to remove stop words from text data in multidimensional data, where stop words refer to words that have no actual semantic value. The text after removing the stop words is segmented, that is, segmented words are extracted from the text after removing the stop words, and only three categories of nouns, verbs, and adjectives are retained in the segmentation results. According to the preset number of topics, the probability of each term in the text data on each topic is counted respectively, and a topic-term distribution is established. The frequency of each term in the text data is counted, and based on the topic-term distribution, a document-topic distribution is established.

[0088] In some embodiments, a sliding window can be used to extract time series features from data containing time series characteristics in multidimensional data, or the relationship between time series characteristic data at each moment can be obtained based on an autoregressive moving average model to extract time series features. Time series features can also be extracted by performing statistical analysis on data containing time series characteristics in multidimensional data.

[0089] According to the embodiments of the present application, by extracting features from multiple dimensions such as text data, numerical data, and data containing time series characteristics, user characteristics and behavior patterns can be accurately portrayed, providing strong support for building accurate user portraits; using the linear discriminant analysis algorithm to extract the topic distribution features of text data can more accurately retain key information. Moreover, when extracting topic distribution features, the linear discriminant analysis algorithm can not only reduce the data dimension, but also maximize the inter-class differences and minimize the intra-class differences, which is conducive to improving the accuracy of user portraits. In addition, the linear discriminant analysis algorithm has low computational complexity and is suitable for processing large-scale, high-dimensional data.

[0090] In some embodiments, the extracted features are fused, including:

[0091] Map each extracted feature to the same dimension;

[0092] The attention mechanism is used to perform weighted fusion of features in the same dimension.

[0093] Furthermore, corresponding fully connected layers are set for numerical features, topic distribution features, and time series features, respectively. Each feature is mapped to the same dimension through its corresponding fully connected layer. For example, the dimension of the extracted numerical features is 20 dimensions, the dimension of the topic distribution features is 30 dimensions, and the dimension of the time series features is 50 dimensions. Each feature can be mapped to 128 dimensions through its corresponding fully connected layer. This example is only an example. The dimensions of each feature after extraction can also be other dimensions, and the same dimension mapped can also be a dimension other than 128 dimensions.

[0094] Furthermore, by constructing an attention function, the attention score of each feature in the same dimension is calculated respectively. According to the attention score of each feature, the weight corresponding to the corresponding feature is determined. Based on the weight corresponding to each feature in the same dimension, the features in the same dimension are weightedly fused.

[0095] According to the embodiments of the present application, mapping each feature to the same dimension can reduce the differences caused by different types of data; weighted fusion of each feature mapped to the same dimension can reduce the problem of low accuracy of user portraits generated based on single-dimensional features, thereby improving the accuracy of user portrait construction.

[0096] Figure 3 The first dimensionality reduction processing flow chart according to an embodiment of the present application is schematically shown.

[0097] like Figure 3 As shown, the first dimensionality reduction process is performed on the fusion features, including:

[0098] Using the uniform manifold approximation projection algorithm, a high-dimensional graph is constructed based on the fusion features;

[0099] Calculate the cross entropy between a high-dimensional graph and a randomly generated low-dimensional graph;

[0100] Optimize the low-dimensional graph according to the cross entropy and replace the randomly generated low-dimensional graph with the optimized low-dimensional graph;

[0101] Repeat the cross entropy calculation and low-dimensional graph optimization operations until the obtained cross entropy is less than or equal to the preset cross entropy threshold;

[0102] The fused features are projected into the first low-dimensional subspace constructed by the low-dimensional graph corresponding to the cross entropy less than or equal to the preset cross entropy threshold.

[0103] For example, assuming that the first preset time is the last year and the second preset time is the last 7 days, feature extraction is performed on the multi-dimensional data in the last year. After obtaining the fused features, the fused features are reduced in dimension using the uniform manifold approximation projection algorithm. This not only reduces the data dimension, but also obtains the user's behavior pattern in the last year, such as the user's quarterly consumption report, the user's topological relationship in the last year, the user's monthly financial management situation, etc. Based on the user's quarterly consumption report, the user's consumption preferences can be learned. Based on the user's topological relationship in the last year, it can be determined whether the user is involved in illegal behavior. Based on the user's monthly financial management situation, the user's preference for financial products can be learned. Based on these long-term behavior patterns, the user's long-term preferences can be effectively obtained, which is beneficial to providing users with corresponding products and improving customer stickiness.

[0104] In some embodiments, constructing a high-dimensional graph based on the fused features includes:

[0105] The dimension of the fused feature is used as the coordinate axis, and the eigenvalues ​​of the data points in the fused feature are used as the coordinate values ​​to construct a high-dimensional space.

[0106] In high-dimensional space, the similarity relationship between each data point in the fusion feature is calculated separately, each data point is regarded as a vertex, and the weight of the edge between the two vertices corresponding to the two data points is determined according to the similarity relationship between the two data points, so as to construct a high-dimensional graph.

[0107] In some embodiments, the cross entropy represents the degree of structural difference between the high-dimensional graph and the low-dimensional graph. The smaller the cross entropy, the closer the low-dimensional graph and the high-dimensional graph match, and the more accurate the high-dimensional graph information embedded in the low-dimensional graph is.

[0108] According to the embodiments of the present application, the uniform manifold approximation projection algorithm can maintain the topological structure of high-dimensional data by continuously optimizing the low-dimensional graph, and then accurately extract the complex correlations between customers' deep behavior patterns (i.e., long-term behavior patterns, such as behavior patterns within the past year, etc.), which is conducive to identifying customers' long-term preference patterns.

[0109] Figure 4 The second dimensionality reduction processing flow chart according to an embodiment of the present application is schematically shown.

[0110] like Figure 4 As shown, a second dimensionality reduction process is performed on the fusion features within the second preset time, including:

[0111] Calculating the covariance matrix of the fusion features within the second preset time by a principal component analysis algorithm;

[0112] Perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and corresponding eigenvectors;

[0113] Sort the eigenvalues ​​in descending order, and construct a principal component matrix based on the eigenvectors corresponding to the first preset number of eigenvalues ​​that are ranked first;

[0114] The fused features within the second preset time are projected into a second low-dimensional subspace formed by the principal component matrix.

[0115] Furthermore, the fusion features within the second preset time can be screened out based on the time series features in the fusion features;

[0116] For example, assuming that the second preset time is the last 5 days and the first preset time is the last 2 years, the fused features within the last 5 days are selected from the fused features within the last 2 years, the covariance matrix of the fused features within the last 5 days is calculated using the principal component analysis algorithm, and the eigenvalue decomposition of the covariance matrix is ​​performed to obtain the eigenvalues ​​and the eigenvectors corresponding to the eigenvalues, the eigenvalues ​​are sorted in descending order, and the principal component matrix is ​​constructed according to the eigenvectors corresponding to the first K eigenvalues, where K is an integer greater than or equal to 1. Finally, the fused features within the last 5 days are projected into the second-dimensional subspace formed by the principal component matrix; wherein the dimension of the second-dimensional subspace may be the same as or different from the dimension of the first-dimensional subspace; combined with the fused features within the last 5 days, the user portrait can be updated in a timely manner.

[0117] According to the embodiments of the present application, the principal component analysis algorithm can maximize the variance direction by constructing a principal component matrix, thereby achieving efficient dimensionality reduction and key information enhancement. The principal component analysis algorithm is used to reduce the dimensionality of the user's recent fusion features, which is conducive to obtaining real-time data features, thereby achieving risk management and timely updating of user portraits.

[0118] In some embodiments, before weighted fusion of the fused features after the first dimensionality reduction processing and the second dimensionality reduction processing, the fused features after the first dimensionality reduction processing and the second dimensionality reduction processing need to be mapped to the same dimension respectively. Mapping the features to the same dimension is conducive to subsequent processing. The same dimension can be the same as the same dimension to which each feature is mapped after extraction, or it can be different. Generally, the same dimension is smaller than the same dimension to which each feature is mapped after extraction.

[0119] Figure 5 The following schematically shows a flowchart of product recommendation based on user portrait according to an embodiment of the present application.

[0120] like Figure 5 As shown, based on the user portrait, corresponding products are recommended to users, including:

[0121] Assess user risk level based on user profile;

[0122] If the user risk level is lower than the preset threshold, obtain the first historical user profile that matches the user profile, and calculate the similarity between the user profile and the product profile corresponding to the first historical user profile;

[0123] If the similarity is greater than the preset similarity threshold, the product corresponding to the product portrait will be recommended to the user.

[0124] Furthermore, the higher the user risk level, the more likely the user is to engage in high-risk behaviors such as potential breach of contract, fraud, or illegal activities;

[0125] The greater the similarity between the user profile and the product profile corresponding to the first historical user profile, the closer the two are to each other. When the two are most closely matched, the product corresponding to the most closely matching product profile is recommended to the user.

[0126] Furthermore, if the similarity is less than or equal to a preset similarity threshold, the product corresponding to the product portrait is taken as the center, and the products within the preset search radius in the preset product library are retrieved as substitutes, and the similarity between the substitute portrait and the user portrait is calculated. If the similarity is still less than or equal to the preset similarity threshold, the substitute is obtained and the similarity between the substitute portrait and the user portrait is calculated repeatedly until the similarity is greater than the preset similarity threshold. For example, when the similarity between the user portrait and the product portrait corresponding to the first historical user portrait is small, it means that the two do not match, and it is necessary to reselect the product to be recommended. The preset product library is searched for products near the product corresponding to the current product portrait according to the preset search radius, and the searched product is used as a substitute. The similarity between the substitute portrait and the user portrait is recalculated, and the above process is repeated until the recommended product that best matches the user is found.

[0127] If the user's risk level is greater than or equal to the preset threshold, the user will be prohibited from handling high-risk businesses. The system may continue to recommend products to the user or no longer recommend products to the user based on the actual situation. For example, if the user's risk level determines that the user has a tendency to engage in high-risk behaviors such as potential default, fraud, or illegal activities, the user will be prohibited from handling loans, installment payments, fund transfers, credit payments and other businesses. For users who have defaulted, the system may recommend some low-risk products to the user based on the user profile, such as books, teaching aids, etc. For users involved in illegal activities, the system may choose not to recommend products to them.

[0128] According to the embodiments of the present application, by screening low-risk customer groups and drawing on the historical choices of historical customers with similar profiles, the accuracy of product recommendations can be improved and the risk of recommended products can be reduced.

[0129] Figure 6 The following schematically shows a product recommendation flow chart for new users according to an embodiment of the present application.

[0130] like Figure 6 As shown, the method 200 further includes:

[0131] For new users, historically recommended products are sorted in descending order of recommendation frequency, and the second preset number of products ranked first are selected as the candidate product set;

[0132] Obtain the second historical user portraits of each product in the candidate product set, calculate the similarity between the new user portrait and the second historical user portrait, and obtain the third historical user that matches the new user;

[0133] Based on the recommended products received by the third historical user, a first recommended product is generated and recommended to the new user.

[0134] Further, if no third historical user is matched, a third preset number of products in the candidate product set are recommended to the new user;

[0135] Generating a first recommended product based on the recommended products received by the third historical user and recommending the first recommended product to the new user, including: sorting the recommended products received by the third historical user according to their frequency of recommendation to the third historical user, selecting the top few products that belong to the candidate product set as the first recommended products and recommending them to the new user;

[0136] For example, for a new user, the products recommended to historical users are sorted in descending order of recommendation frequency, and the top M products are selected as the candidate product set, where M is an integer greater than or equal to 1; the second historical user portrait involved in each of the top M products is obtained, for example, if product 1 has been recommended to historical user A and historical user B, then the second historical user portrait involved in product 1 includes the portrait of historical user A and the portrait of historical user B. For another example, if historical user C has used product 1, then the second historical user portrait involved in product 1 should also include the portrait of historical user C. The similarity between the new user portrait and each second historical user portrait is calculated respectively, and the historical user corresponding to the second historical user portrait with the greatest similarity is selected as the third historical user. The products recommended to the third historical user are sorted in descending order of recommendation frequency, and the top Q products are selected as the first recommended products to be recommended to the new user, where Q is an integer greater than or equal to 1;

[0137] If no third historical user is matched, the first Y products are selected from the first M products and recommended to the new user, where Y is an integer greater than or equal to 1.

[0138] According to an embodiment of the present application, based on the recommendation frequency of historically recommended products, popular products can be recommended to new users; combined with the matching historical products previously received by users, personalized recommendations can be implemented for new users.

[0139] In some embodiments, the user portrait can be updated in a timely manner based on the recent multi-dimensional data obtained, and the product recommendation algorithm can also be optimized based on user feedback and the recent multi-dimensional data obtained.

[0140] In some embodiments, method 200 can be applied to multiple scenarios such as accurate identification of high-net-worth customers, credit card product recommendations, loan risk assessment, mobile banking user behavior analysis, and cross-channel collaboration. For example, high-net-worth customers can be identified based on user portraits and high-end services such as private banking can be recommended to them; based on user portraits, airport VIP lounge benefits can be recommended to business travelers, and educational products can be recommended to users belonging to the "parent-child family" category to improve the accuracy of recommendations; user risk levels can be assessed based on user portraits to obtain user loan risks. For enterprises whose portrait features include "high-prosperity industry + low supply chain", even if the enterprise's current profit is low, it belongs to a high-prosperity industry, so loan products can still be recommended to them; in mobile banking user behavior analysis, for users whose portraits show "high dependence + low proficiency", a novice guidance window is provided to the user, and for users whose portraits show "low sensitivity + high-frequency operation", a quick function customization window is recommended to the user; based on user portraits, corresponding products can be recommended to users through various forms such as mobile banking, bank customer service systems, and offline outlets, thereby achieving cross-channel collaboration.

[0141] Based on the above-mentioned personalized recommendation method based on user portrait, this application also provides a personalized recommendation device based on user portrait. Figure 7 The device is described in detail.

[0142] Figure 7 The structural block diagram of the personalized recommendation device based on user portrait according to an embodiment of the present application is schematically shown.

[0143] like Figure 7 As shown, the user portrait-based personalized recommendation device 700 of this embodiment includes a multi-dimensional data acquisition module 710, a fusion feature acquisition module 720, a dimensionality reduction processing module 730, a user portrait acquisition module 740 and a product recommendation module 750.

[0144] Multi-dimensional data acquisition module 710 is configured to acquire multi-dimensional data about a user within a first preset time period; the multi-dimensional data includes basic user information, behavioral data, time data, and topological relationships. In one embodiment, multi-dimensional data acquisition module 710 can be configured to perform operation S210 described above and will not be further described here.

[0145] The fusion feature acquisition module 720 is used to extract features from the multi-dimensional data and fuse the extracted features to obtain fusion features. In one embodiment, the fusion feature acquisition module 720 can be used to perform the operation S220 described above, which will not be repeated here.

[0146] Dimensionality reduction processing module 730 is configured to perform a first dimensionality reduction process on the fused features and a second dimensionality reduction process on the fused features within a second preset time period; wherein the first preset time period is within the second preset time period and the interval between the second preset time period and the current time period is less than or equal to a preset time threshold. In one embodiment, dimensionality reduction processing module 730 can be configured to perform operation S230 described above and will not be further described here.

[0147] The user portrait acquisition module 740 is configured to perform weighted fusion on the fused features after the first dimensionality reduction process and the second dimensionality reduction process to obtain a user portrait. In one embodiment, the user portrait acquisition module 740 may be configured to perform the operation S230 described above, which will not be described in detail here.

[0148] The product recommendation module 750 is configured to recommend corresponding products to the user based on the user portrait. In one embodiment, the product recommendation module 750 may be configured to perform the operation S230 described above, which will not be described in detail herein.

[0149] In some embodiments, the fusion feature acquisition module 720 is specifically used to:

[0150] Feature extraction for multi-dimensional data, including:

[0151] Standardize the numerical data in the multi-dimensional data to obtain standardized numerical features;

[0152] Use the linear discriminant analysis algorithm to perform topic analysis on text data in multi-dimensional data and extract the topic distribution characteristics of text data;

[0153] Time series features are extracted from multi-dimensional data containing time series characteristics to obtain time series features.

[0154] In some embodiments, the fusion feature acquisition module 720 is further configured to:

[0155] The features extracted after fusion include:

[0156] Mapping the extracted features to the same dimension respectively;

[0157] The attention mechanism is used to perform weighted fusion of the features in the same dimension.

[0158] In some embodiments, the dimensionality reduction processing module 730 is specifically used to:

[0159] Performing a first dimensionality reduction process on the fused features, including:

[0160] Using a uniform manifold approximation projection algorithm, constructing a high-dimensional graph according to the fused features;

[0161] Calculating the cross entropy between the high-dimensional graph and a randomly generated low-dimensional graph;

[0162] Optimizing the low-dimensional graph according to the cross entropy, and replacing the randomly generated low-dimensional graph with the optimized low-dimensional graph;

[0163] Repeat the cross entropy calculation and low-dimensional graph optimization operations until the obtained cross entropy is less than or equal to the preset cross entropy threshold;

[0164] The fused features are projected into a first low-dimensional subspace constructed by a low-dimensional graph corresponding to a cross entropy that is less than or equal to a preset cross entropy threshold.

[0165] In some embodiments, the dimensionality reduction processing module 730 is further configured to:

[0166] Performing a second dimensionality reduction process on the fusion features within the second preset time includes:

[0167] Calculating the covariance matrix of the fusion features within the second preset time by a principal component analysis algorithm;

[0168] Performing eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and corresponding eigenvectors;

[0169] Sorting the eigenvalues ​​in descending order, and constructing a principal component matrix based on the eigenvectors corresponding to the first preset number of eigenvalues ​​that are ranked first;

[0170] The fused features within the second preset time are projected into a second low-dimensional subspace formed by the principal component matrix.

[0171] In some embodiments, the product recommendation module 750 is specifically configured to:

[0172] Based on the user profile, recommend corresponding products to the user, including:

[0173] Assessing the user's risk level based on the user profile;

[0174] If the user risk level is lower than a preset threshold, obtaining a first historical user profile that matches the user profile, and calculating the similarity between the user profile and the product profile corresponding to the first historical user profile;

[0175] If the similarity is greater than a preset similarity threshold, the product corresponding to the product portrait is recommended to the user.

[0176] In some embodiments, the apparatus 700 is further configured to:

[0177] For new users, historically recommended products are sorted in descending order of recommendation frequency, and the second preset number of products ranked first are selected as the candidate product set;

[0178] Obtaining the second historical user portraits associated with each product in the candidate product set, respectively, calculating the similarity between the new user portrait and the second historical user portrait, and obtaining a third historical user that matches the new user;

[0179] Based on the recommended products received by the third historical user, a first recommended product is generated and recommended to the new user.

[0180] According to an embodiment of the present application, the device 700 can improve the accuracy of user portrait construction, thereby improving the recommendation accuracy and user experience.

[0181] According to an embodiment of the present application, any multiple modules among the multi-dimensional data acquisition module 710, the fused feature acquisition module 720, the dimensionality reduction processing module 730, the user portrait acquisition module 740, and the product recommendation module 750 can be combined into a single module, or any one of them can be split into multiple modules. Alternatively, at least part of the functionality of one or more of these modules can be combined with at least part of the functionality of other modules and implemented in a single module. According to an embodiment of the present application, at least one of the multi-dimensional data acquisition module 710, the fused feature acquisition module 720, the dimensionality reduction processing module 730, the user portrait acquisition module 740, and the product recommendation module 750 can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on a chip, a system on a substrate, a system on a package, an application-specific integrated circuit (ASIC), or can be implemented in hardware or firmware by any other reasonable means of integrating or packaging circuits, or can be implemented in any one of the three implementation methods of software, hardware, and firmware, or in an appropriate combination of any of them. Alternatively, at least one of the multi-dimensional data acquisition module 710, the fusion feature acquisition module 720, the dimensionality reduction processing module 730, the user portrait acquisition module 740 and the product recommendation module 750 can be at least partially implemented as a computer program module, which can perform corresponding functions when the computer program module is run.

[0182] Figure 8 A block diagram of an electronic device suitable for implementing a personalized recommendation method based on user portraits according to an embodiment of the present application is schematically shown.

[0183] like Figure 8 As shown, an electronic device 800 according to an embodiment of the present application includes a processor 801, which can perform various appropriate actions and processes based on a program stored in a read-only memory (ROM) 802 or a program loaded from a storage unit 808 into a random access memory (RAM) 803. The processor 801 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or a related chipset and / or a dedicated microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 801 may also include onboard memory for caching purposes. The processor 801 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present application.

[0184] Various programs and data required for the operation of the electronic device 800 are stored in the RAM 803. The processor 801, ROM 802, and RAM 803 are connected to each other via a bus 804. The processor 801 performs various operations of the method flow according to the embodiment of the present application by executing the programs in the ROM 802 and / or RAM 803. It should be noted that the programs may also be stored in one or more memories other than the ROM 802 and the RAM 803. The processor 801 may also perform various operations of the method flow according to the embodiment of the present application by executing the programs stored in the one or more memories.

[0185] According to an embodiment of the present application, electronic device 800 may further include an input / output (I / O) interface 805, which is also connected to bus 804. Electronic device 800 may also include one or more of the following components connected to I / O interface 805: an input section 806 including a keyboard, mouse, etc.; an output section 807 including devices such as a cathode ray tube (CRT), liquid crystal display (LCD), and speakers; a storage section 808 including a hard disk; and a communication section 808 including a network interface card such as a LAN card or modem. Communication section 808 performs communication processing via a network such as the Internet. A drive 810 is also connected to I / O interface 805 as needed. Removable media 811, such as a magnetic disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed in drive 810 as needed, so that computer programs read from the removable media can be installed into storage section 808 as needed.

[0186] This application also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments, or may exist independently and not be incorporated into the device / apparatus / system. The computer-readable storage medium carries one or more programs, and when the one or more programs are executed, the method according to the embodiments of this application is implemented.

[0187] According to an embodiment of the present application, a computer-readable storage medium may be a non-volatile computer-readable storage medium, and may include, for example, but not limited to: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present application, a computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, according to an embodiment of the present application, a computer-readable storage medium may include the ROM 802 and / or RAM 803 described above and / or one or more memories other than ROM 802 and RAM 803.

[0188] The embodiments of the present application also include a computer program product, which includes a computer program containing program code for executing the method shown in the flowchart. When the computer program product is executed in a computer system, the program code is used to enable the computer system to implement the personalized recommendation method based on user profiles provided in the embodiments of the present application.

[0189] The computer program executes the above functions defined in the system / device of the embodiment of the present application when the processor 801 executes the computer program. According to the embodiment of the present application, the system, device, module, unit, etc. described above can be implemented by a computer program module.

[0190] In one embodiment, the computer program may be stored on a tangible storage medium such as an optical storage device or a magnetic storage device. In another embodiment, the computer program may be transmitted and distributed in the form of a signal on a network medium, downloaded and installed via the communication portion 808, and / or installed from a removable medium 811. The program code contained in the computer program may be transmitted using any appropriate network medium, including but not limited to wireless, wired, or any suitable combination thereof.

[0191] In such an embodiment, the computer program can be downloaded and installed from a network via the communication section 808, and / or installed from a removable medium 811. When the computer program is executed by the processor 801, the above-described functions defined in the system of the embodiment of the present application are performed. According to the embodiment of the present application, the systems, devices, means, modules, units, etc. described above can be implemented by computer program modules.

[0192] According to an embodiment of the present application, the program code for executing the computer program provided by the embodiment of the present application can be written in any combination of one or more programming languages. Specifically, these computer programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, languages ​​such as Java, C++, Python, "C" or similar programming languages. The program code can be executed entirely on the user computing device, partially on the user device, partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (for example, using an Internet service provider to connect via the Internet).

[0193] The flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, program segment, or a part of code, and the above-mentioned module, program segment, or a part of code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of the boxes in the block diagram or flowchart, can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0194] Those skilled in the art will appreciate that the features described in the various embodiments of this application may be combined and / or coupled in various ways, even if such combinations or couplings are not explicitly described in this application. In particular, the features described in the various embodiments of this application may be combined and / or coupled in various ways without departing from the spirit and teachings of this application. All such combinations and / or couplings fall within the scope of this application.

Claims

1. A personalized recommendation method based on user portrait, characterized in that: The method comprises: Acquire multi-dimensional data of the user within a first preset time; wherein the multi-dimensional data includes basic user information, behavior data, time data and topological relationships; Performing feature extraction on the multi-dimensional data, and fusing the extracted features to obtain fused features; Performing a first dimensionality reduction process on the fused features, and performing a second dimensionality reduction process on the fused features within a second preset time; wherein the first preset time is within the second preset time and the interval between the second preset time and the current time is less than or equal to a preset time threshold; Performing weighted fusion on the fused features after the first dimensionality reduction processing and the second dimensionality reduction processing to obtain a user profile; Based on the user portrait, corresponding products are recommended to the user.

2. The method according to claim 1, characterized in that The extracting features from the multi-dimensional data includes: Normalizing the numerical data in the multi-dimensional data to obtain standardized numerical features; Performing topic analysis on the text data in the multidimensional data using a linear discriminant analysis algorithm to extract topic distribution features of the text data; Time series features are extracted from the data containing time series characteristics in the multi-dimensional data to obtain time series features.

3. The method according to claim 1, characterized in that The features extracted after fusion include: Mapping the extracted features to the same dimension respectively; The attention mechanism is used to perform weighted fusion of the features in the same dimension.

4. The method according to claim 1, wherein The performing a first dimensionality reduction process on the fusion feature includes: Using a uniform manifold approximation projection algorithm, constructing a high-dimensional graph according to the fused features; Calculating the cross entropy between the high-dimensional graph and a randomly generated low-dimensional graph; Optimizing the low-dimensional graph according to the cross entropy, and replacing the randomly generated low-dimensional graph with the optimized low-dimensional graph; Repeat the cross entropy calculation and low-dimensional graph optimization operations until the obtained cross entropy is less than or equal to the preset cross entropy threshold; The fused features are projected into a first low-dimensional subspace constructed by a low-dimensional graph corresponding to a cross entropy that is less than or equal to a preset cross entropy threshold.

5. The method according to claim 1, wherein The performing a second dimensionality reduction process on the fusion features within the second preset time includes: Calculating the covariance matrix of the fusion features within the second preset time by a principal component analysis algorithm; Performing eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and corresponding eigenvectors; Sorting the eigenvalues ​​in descending order, and constructing a principal component matrix based on the eigenvectors corresponding to the first preset number of eigenvalues ​​that are ranked first; The fused features within the second preset time are projected into a second low-dimensional subspace formed by the principal component matrix.

6. The method according to claim 1, wherein The recommending corresponding products to the user based on the user portrait includes: Assessing the user's risk level based on the user profile; If the user risk level is lower than a preset threshold, obtaining a first historical user profile that matches the user profile, and calculating the similarity between the user profile and the product profile corresponding to the first historical user profile; If the similarity is greater than a preset similarity threshold, the product corresponding to the product portrait is recommended to the user.

7. The method according to claim 1, characterized in that The method further comprises: For new users, historically recommended products are sorted in descending order of recommendation frequency, and the second preset number of products ranked first are selected as the candidate product set; Obtaining the second historical user portraits associated with each product in the candidate product set, respectively, calculating the similarity between the new user portrait and the second historical user portrait, and obtaining a third historical user that matches the new user; Based on the recommended products received by the third historical user, a first recommended product is generated and recommended to the new user.

8. A personalized recommendation device based on user portrait, characterized in that: The device comprises: A multi-dimensional data acquisition module is used to acquire multi-dimensional data of the user within a first preset time; wherein the multi-dimensional data includes basic user information, behavior data, time data and topological relationships; A fusion feature acquisition module is used to extract features from the multi-dimensional data and fuse the extracted features to obtain fusion features; a dimensionality reduction processing module, configured to perform a first dimensionality reduction processing on the fused features, and perform a second dimensionality reduction processing on the fused features within a second preset time; wherein the first preset time is included in the second preset time and the interval between the second preset time and the current time is less than or equal to a preset time threshold; A user portrait acquisition module, configured to perform weighted fusion on the fused features after the first dimensionality reduction processing and the second dimensionality reduction processing to obtain a user portrait; The product recommendation module is used to recommend corresponding products to users based on the user portrait.

9. An electronic device comprising: one or more processors; a memory for storing one or more computer programs, It is characterized in that the one or more processors execute the one or more computer programs to implement the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program or instruction stored thereon, characterized in that: When the computer program or instruction is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

11. A computer program product comprising a computer program or instructions, characterized in that When the computer program or instruction is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images