Financial service pushing method and device and electronic equipment

Abstract

The invention provides a financial service pushing method and device and electronic equipment. Comprising the steps that user data are acquired, user features are extracted according to the user data, the user data comprise user identity information and user consumption information, and the user features represent consumption features of a user; the user features are analyzed through a user portrait model, a user category prediction result is obtained, the user category prediction result represents the consumption level of the user, and the user portrait model is obtained through training of multiple groups of data; each group of data in the multiple groups of data comprises historical user features and user category prediction results corresponding to the historical user features; a user portrait is generated according to the user category prediction result, service pushing is carried out on the user according to the user portrait, and the user portrait at least comprises consumption characteristics of the user. Through the method and the device, the problem of poor user experience caused by the fact that financial services cannot be accurately pushed in the prior art is solved.

Description

Technical Field

[0001] The present application relates to the technical field of financial business push, and more specifically, to a method, device, computer-readable storage medium, and electronic device for pushing financial business. Background Art

[0002] With the popularization of mobile payment and the change of the public's consumption concept, the business promotion mode of financial institutions has changed from "pull type" to "push type". Merchants can achieve refined customer management through big data analysis. By optimizing user portraits, they can accurately push preferential mechanisms to consumers to achieve accurate recommendation, thereby improving the promotion efficiency of financial business and reducing the promotion cost. At the same time, for consumers, they can also enjoy personalized services and discounts when purchasing goods, enhancing the consumption experience. However, due to the complexity of user portraits, the real-time nature of data, and the uncertainty of user behavior, the above-mentioned accurate promotion method for financial business still has the following problems in practical applications:

[0003] 1. In actual financial business promotion activities, due to the influence of factors such as time and space on data collection, the real-time nature of data is poor, resulting in difficulty in updating user portraits in a timely manner. The update cycle of most customer portraits is one year or longer, and it is impossible to achieve real-time update of user portraits, which cannot meet the rapidly changing promotion strategies and customer needs, resulting in the failure of the issuance of preferential mechanisms to achieve the expected effect.

[0004] 2. When the update cycle of the user portrait is relatively long, user tags and features will change over time, but most models assume that the tags and features are static during the training process and cannot adapt to the scenario of dynamic changes in tags and features, resulting in poor accuracy of model prediction results.

[0005] In summary, the existing promotion methods for financial business cannot perform accurate push, resulting in poor user experience. Summary of the Invention

[0006] The main purpose of the present application is to provide a method, device, computer-readable storage medium, and electronic device for pushing financial business, so as to at least solve the problem that the existing technology cannot perform accurate push of financial business, resulting in poor user experience.

[0007] To achieve the above object, according to one aspect of the present application, there is provided a method for pushing financial services, including: obtaining user data, extracting user features according to the user data, where the user data includes user identity information and user consumption information, and the user features characterize the consumption characteristics of the user; analyzing the user features through a user portrait model to obtain a user category prediction result, where the user category prediction result represents the consumption level of the user, and the user portrait model is trained through multiple sets of data, and each set of data in the multiple sets of data includes historical user features and the user category prediction result corresponding to the historical user features; generating a user portrait according to the user category prediction result, and pushing services to the user according to the user portrait, where the user portrait at least includes the consumption characteristics of the user.

[0008] Optionally, before extracting user features according to the user data, the method further includes: labeling the user data to obtain static labels and dynamic labels, where the static labels represent the unchanging user data, and the dynamic labels represent the dynamically changing user data; performing a normalization process on the user data labeled with the static labels and the dynamic labels to obtain the normalized user data.

[0009] Optionally, extracting user features according to the user data includes: determining the value of the user data through a value evaluation model; obtaining a one-to-one mapping relationship between the value and preset user features, and determining the preset user features corresponding to the value of the user data according to the one-to-one mapping relationship to obtain the user features.

[0010] Optionally, pushing services to the user according to the user portrait includes: obtaining the target push products, user geographical location, user consumption frequency, user consumption level, and user interaction times corresponding to the user portrait, where the target push products represent business products with the user's purchase times greater than a first preset threshold, and the user interaction times represent the number of times the user participates in historical push activities; determining a push mechanism according to the target push products, the user geographical location, the user consumption frequency, the user consumption level, and the user interaction times, and pushing business products to the user according to the push mechanism.

[0011] Optionally, a push mechanism is determined according to the target push product, the user's geographical location, the user's consumption frequency, the user's consumption level, and the number of user interactions, including: pushing the target push product to the user again; obtaining the business product address, calculating the absolute value of the difference between the business product address and the user's geographical location, and pushing the business product corresponding to the business product address whose absolute value of the difference is less than a second preset threshold to the user; and pushing a preferential mechanism to the user when the user's consumption frequency is greater than a third preset threshold or the user's consumption level is greater than a fourth preset threshold or the number of user interactions is greater than a fifth preset threshold.

[0012] Optionally, before analyzing the user characteristics through the user portrait model, the method further includes: constructing an initial user portrait model and obtaining the historical user characteristics, where the initial user portrait model includes an input layer, a hidden layer, and an output layer; inputting the historical user characteristics into the input layer of the initial user portrait model, and analyzing the historical user characteristics through the hidden layer until the loss function is minimized to obtain the user portrait model.

[0013] Optionally, analyzing the historical user characteristics through the hidden layer includes: passing the historical user characteristics through the function h (l) = σ(W (l) h (l-1) + b (l) ) to analyze the historical user characteristics, where σ represents an activation function, W (l) represents the weight of the l-th layer, b (l) represents the bias of the l-th layer, h (l-1) represents the output result of the hidden layer of the (l - 1)-th layer, and h (l) represents the output result of the hidden layer of the l-th layer.

[0014] According to another aspect of the present application, there is provided a push device for financial services, including: an extraction unit for obtaining user data and extracting user characteristics according to the user data, where the user data includes user identity information and user consumption information, and the user characteristics characterize the consumption characteristics of the user; a first analysis unit for analyzing the user characteristics through a user portrait model to obtain a user category prediction result, where the user category prediction result represents the consumption level of the user, and the user portrait model is trained through multiple sets of data, and each set of data in the multiple sets of data includes historical user characteristics and the user category prediction result corresponding to the historical user characteristics; and a push unit for generating a user portrait according to the user category prediction result and performing business push on the user according to the user portrait, where the user portrait at least includes the consumption characteristics of the user.

[0015] According to another aspect of the present application, there is provided a computer-readable storage medium, which includes a stored program. When the program runs, it controls the device where the computer-readable storage medium is located to execute any one of the above-mentioned financial service push methods.

[0016] According to yet another aspect of the present application, there is provided an electronic device, including: one or more processors, a memory, and one or more programs. The one or more programs are stored in the memory and are configured to be executed by the one or more processors. The one or more programs include a method for executing any one of the above-mentioned financial service push methods.

[0017] Applying the technical solution of the present application, user data is obtained, and user characteristics are extracted based on the user data. The user data includes user identity information and user consumption information, and the user characteristics represent the consumption characteristics of the user. The user characteristics are analyzed through a user portrait model to obtain a user category prediction result. The user category prediction result represents the consumption level of the user. The user portrait model is trained through multiple groups of data, and each group of data in the multiple groups of data includes historical user characteristics and the user category prediction result corresponding to the historical user characteristics. A user portrait is generated according to the user category prediction result, and business push is performed on the user according to the user portrait. The user portrait at least includes the consumption characteristics of the user. Compared with the prior art where financial services cannot be accurately pushed, in the present application, user characteristics are extracted to generate a user portrait, and business push is performed according to the user portrait, so that accurate push can be performed according to the consumption characteristics of the user, improving the user experience. Therefore, the problem of inaccurate push in the prior art can be solved, and the effect of accurately pushing financial services to relevant users can be achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The accompanying drawings forming a part of this application are used to provide a further understanding of this application. The schematic embodiments of this application and their descriptions are used to explain this application and do not constitute an improper limitation to this application. In the drawings:

[0019] Figure 1 It shows a hardware structure block diagram of a mobile terminal for executing a financial service push method provided by an embodiment of this application;

[0020] Figure 2 It shows a schematic flowchart of a financial service push method provided by an embodiment of this application;

[0021] Figure 3 It shows a schematic flowchart of a specific financial service push method provided by an embodiment of this application;

[0022] Figure 4The block diagram of a push device for a financial service provided by an embodiment of the present application is shown.

[0023] Among them, the above-mentioned drawings include the following reference numerals:

[0024] 102. Processor; 104. Memory; 106. Transmission device; 108. Input / output device. Detailed implementation manners

[0025] It should be noted that, without conflict, the embodiments in the present application and the features in the embodiments may be combined with each other. The present application will be described in detail below with reference to the drawings and in combination with the embodiments.

[0026] In order to enable those skilled in the art to better understand the solution of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0027] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so as to describe the embodiments of the present application here. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device including a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0028] For the convenience of description, some nouns or terms related to the embodiments of the present application are described below:

[0029] User portrait: A way to deeply analyze and describe a user's personal information, consumption behavior, hobbies, etc. through technical means such as data analysis and mining, so as to form a complete user image.

[0030] Preference mechanism push: Refers to providing a customized in-store consumption preference mechanism for different users according to factors such as their basic information and consumption habits.

[0031] RFM model: A customer classification model used to describe the customer value situation, which is a data analysis model for measuring customer value and customer profit-making ability. It describes the customer value situation through three key dimensions: Recency (the time of the last consumption), Frequency (the consumption frequency), and Monetary (the consumption amount).

[0032] MLP algorithm: Multilayer Perceptron, abbreviated as MLP, is a basic feedforward artificial neural network composed of multiple neurons with multiple hidden layers. It learns complex patterns and feature representations of input data through multiple non-linear transformations and is commonly used to solve problems such as classification and regression.

[0033] As introduced in the background art, in the prior art, the inability to perform accurate push in financial services leads to poor user experience. To solve the problem of poor user experience caused by the inability to perform accurate push in financial services, the embodiments of the present application provide a push method, device, computer-readable storage medium, and electronic device for financial services.

[0034] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention.

[0035] The method embodiments provided in the embodiments of the present application can be executed on a mobile terminal, a computer terminal, or a similar computing device. Taking running on a mobile terminal as an example, Figure 1 is a hardware structure block diagram of a mobile terminal for a push method of financial services in the embodiments of the present invention. As Figure 1 shown, the mobile terminal may include one or more ( Figure 1 only one is shown in the figure) processors 102 (the processor 102 may include, but is not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory 104 for storing data. Among them, the above mobile terminal may further include a transmission device 106 for communication functions and an input / output device 108. Those of ordinary skill in the art can understand that Figure 1 the structure shown is only schematic and does not limit the structure of the above mobile terminal. For example, the mobile terminal may further include more or fewer components than Figure 1 shown in the figure, or have a different configuration from Figure 1 shown in the figure.

[0036] The memory 104 can be used to store computer programs, such as software programs and modules of application software, such as the computer program corresponding to the financial service push method in the embodiments of the present invention. The processor 102 executes various functional applications and data processing by running the computer programs stored in the memory 104, that is, implements the above-mentioned method. The memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memories. In some instances, the memory 104 may further include a memory remotely disposed relative to the processor 102, and these remote memories can be connected to the mobile terminal through a network. Examples of the above-mentioned network include but are not limited to the Internet, enterprise intranet, local area network, mobile communication network, and combinations thereof. The transmission device 106 is used to receive or send data via a network. Specific examples of the above-mentioned network may include the wireless network provided by the communication provider of the mobile terminal. In one instance, the transmission device 106 includes a network adapter (Network Interface Controller, abbreviated as NIC), which can be connected to other network devices through a base station and thus can communicate with the Internet. In one instance, the transmission device 106 may be a radio frequency (Radio Frequency, abbreviated as RF) module, which is used to communicate with the Internet wirelessly.

[0037] In this embodiment, a financial service push method running on a mobile terminal, a computer terminal, or a similar computing device is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0038] Figure 2 It is a flowchart of the financial service push method according to the embodiments of the present application. As Figure 2 shown, the method includes the following steps:

[0039] Step S201, obtain user data, and extract user features according to the user data, where the user data includes user identity information and user consumption information, and the user features characterize the consumption features of the user;

[0040] Specifically, through user authorization, personal information data and consumption habit data of users are collected. Obtaining user data and extracting user features is the first and crucial step in building a user portrait. User data can be collected from multiple channels, including but not limited to: User identity information: including basic information such as the user's age, gender, occupation, educational background, marital status, etc. These information help to understand the background characteristics of users, so as to infer their possible consumption preferences and behavior patterns. User consumption information: covering the user's consumption records, including consumption time, consumption location, consumption amount, consumption category, payment method, etc. These data reflect the actual consumption behavior and consumption habits of users. User features extracted from these data can include: consumption preferences, consumption frequency, consumption amount, geographical location, user engagement. When extracting user features, the following technical means can be adopted: data cleaning, feature engineering, and then through machine learning models: using deep learning models such as multi-layer perceptron (MLP) or RFM models, etc., to analyze user data, extract key features, and build a user portrait.

[0041] Step S202, analyze the user features through the user portrait model to obtain a user category prediction result, where the user category prediction result represents the consumption level of the user, and the user portrait model is trained through multiple sets of data, and each set of data in the multiple sets of data includes historical user features and the user category prediction result corresponding to the historical user features;

[0042] Specifically, the process of analyzing user features through the user portrait model to obtain a user category prediction result is an advanced data analysis based on deep learning technology (such as multi-layer perceptron, MLP). This process can be divided into several key steps: Data collection and preprocessing: First, a large amount of user data needs to be collected, including user identity information, consumption information, etc., to form historical user features. Preprocessing includes data cleaning, missing value processing, feature selection, feature encoding, etc., to ensure data quality and model training effect. Feature and label preparation: For each set of historical data, extract user features (such as age, gender, consumption amount, consumption frequency, etc.) and label the consumption level of the user (such as high-level, middle-level, low-level consumers). The label of the consumption level is usually obtained based on the comprehensive analysis of the user's consumption behavior and consumption amount. Model training: Use the multi-layer perceptron model for training. The multi-layer perceptron is a neural network for supervised learning that can learn the complex relationship between input features and output labels. In the training stage, the model will predict the consumption level based on the input user features, and continuously adjust the weights and biases of the model through a loss function (such as cross-entropy loss) and backpropagation algorithm, so that the prediction result is as close as possible to the actual label.

[0043] In this process, the key lies in training a model through a large amount of historical data that can accurately predict the user's consumption level, so as to help financial institutions better understand and predict user behavior, and formulate more effective preferential strategies. In this way, financial institutions can improve the push efficiency, reduce ineffective delivery, and enhance customer satisfaction and loyalty. At the same time, based on the prediction and classification of user behavior, financial institutions can also develop more customized products and services to meet the needs of customers at different consumption levels, thereby improving the market competitiveness and customer service level of financial institutions.

[0044] Step S203: Generate a user portrait according to the user category prediction result, and perform business push on the user according to the user portrait, where the user portrait at least includes the consumption characteristics of the user.

[0045] Specifically, when analyzing user characteristics through a user portrait model (such as an MLP model), user category prediction results are obtained. These prediction results can be a combination of labels such as consumption level, consumption preference, and consumption frequency, so as to construct a complete user portrait. The generation of a user portrait is not just a simple aggregation of prediction results, but also needs to combine the prediction results with other information of the user (such as basic information, geographical location, hobbies, etc.) to form a more comprehensive and in-depth understanding of the user. The user portrait can include, but is not limited to, the following information: Consumption characteristics: including consumption preferences, consumption frequency, consumption amount, etc. User basic information: age, gender, occupation, educational background, marital status, etc. Geographical location: frequently visited places, place of residence, place of work, etc. User behavior characteristics: frequency of participating in activities, feedback, purchase channel preferences, etc. With the user portrait, financial institutions can formulate more accurate business push strategies based on the consumption characteristics and preferences of the users in the portrait.

[0046] In this embodiment, user data is obtained, and user characteristics are extracted according to the user data. The user data includes user identity information and user consumption information, and the user characteristics represent the consumption characteristics of the user; the user characteristics are analyzed through a user portrait model to obtain user category prediction results, where the user category prediction results represent the consumption level of the user, and the user portrait model is trained through multiple sets of data. Each set of data in the multiple sets of data includes historical user characteristics and user category prediction results corresponding to the historical user characteristics; a user portrait is generated according to the user category prediction result, and business push is performed on the user according to the user portrait, where the user portrait at least includes the consumption characteristics of the user. Compared with the prior art where financial services cannot be accurately pushed, in this application, user characteristics are extracted to generate a user portrait, and business push is performed according to the user portrait, so that accurate push can be performed according to the consumption characteristics of the user, improving the user experience. Therefore, the problem of inaccurate push in the prior art can be solved, and the effect of accurately pushing financial services to relevant users can be achieved.

[0047] In the specific implementation process, before extracting user features based on the user data, the method further includes: Step S204: Label the user data to obtain static labels and dynamic labels, where the static labels represent the unchanging user data, and the dynamic labels represent the user data that changes dynamically; Step S205: Perform standardization processing on the user data labeled with the static labels and the dynamic labels to obtain the standardized user data. The method performs standardization processing, and the data after standardization processing helps the model better identify and learn the consumption characteristics of users, avoiding model skewness caused by differences in data dimensions or ranges. The real-time update of the dynamic labels ensures the timeliness and accuracy of the user portrait, enabling financial institutions to adjust preferential policies in a timely manner to adapt to changes in user behavior.

[0048] Specifically, data annotation is the process of converting raw data into a machine-readable and processable format. At this stage, the collected user data is analyzed to distinguish which data is static and which data changes dynamically. Static labels: These labels represent the basic unchanging information of users, such as age, gender, occupation, educational background, marital status, etc. These information usually do not change frequently and are the basis for constructing the user portrait. Dynamic labels: In contrast to static labels, dynamic labels record the behavioral information of users that changes over time, such as recent consumption records, activity participation, interaction feedback, etc. Dynamic labels are crucial for capturing the current preferences and behavioral patterns of users. Data standardization processing: Data standardization processing is the process of converting user data to the same scale, aiming to eliminate the differences in dimensions and numerical ranges between different features, enabling the model to learn more effectively. Standardization processing usually includes the following steps: Numerical conversion: Convert non-numerical data (such as gender, occupation) into numerical data for easy model processing. Common conversion methods include one-hot encoding and label encoding. Data normalization: Scale numerical data to a fixed range, usually the [0,1] interval, to avoid the influence of numerical differences between features on model training. Normalization methods include min-max scaling and z-score standardization. Missing value processing: Check and process missing values in the data, and methods such as deletion, filling with the mean, and predictive filling can be used to ensure the integrity of the data and the accuracy of model training.

[0049] In some alternative embodiments, the above step S201 of extracting user features from the user data can be implemented through the following steps: Step S2011: Determine the value of the user data through a value evaluation model; Step S2012: Obtain the one-to-one mapping relationship between the value and the preset user features, and determine the preset user features corresponding to the value of the user data according to the one-to-one mapping relationship to obtain the user features. Through the above steps, this method can convert the value of user data into a deep understanding of user features, and then provide more accurate and personalized services and offers for different user groups. At the same time, it can effectively identify and control risks, and improve the operational efficiency and customer satisfaction of the overall business.

[0050] In the specific implementation process, a set of value evaluation models need to be established first. This model can quantify the value of user data for the business. The value evaluation model may be based on multiple factors, such as the consumption ability, loyalty, activity, and potential risks of users. The value evaluation can be performed through the RFM model. The RFM model is a customer classification model used to describe the customer value status and is a data analysis model for measuring customer value and customer profit-making ability. It describes the customer value status through three key dimensions: the time of the last consumption (Recency), the consumption frequency (Frequency), and the consumption amount (Monetary). The preset user features refer to a series of tags or indicators predefined by financial institutions to describe user attributes and behaviors. Before establishing the mapping relationship with the value, the financial institution needs to define which user features correspond to high-value, medium-value, or low-value user data. For example, users with high consumption frequency, high consumption amount, and active participation in financial institution activities may be defined as high-value, while users with low consumption and less participation may be defined as low-value. Once the value evaluation model gives the value of the user data, the financial institution can determine the user features corresponding to the value of the user data according to the preset value feature mapping relationship. For example, if a certain user data is evaluated as high-value, the financial institution may determine that the user has features such as high consumption ability, high usage frequency, or high loyalty. According to the obtained user features, the financial institution can customize service strategies, such as launching high-end product or service offers for users with high consumption ability, and increasing integral rewards or membership upgrade opportunities for users with high activity. In addition, the financial institution can also identify potential business risks, such as fraud risks and credit risks, through user features, and manage risks in a more efficient manner.

[0051] In some alternative embodiments, the above step S203 of pushing services to the user according to the user profile can be implemented through the following steps: Step S2031: Obtain the target push product, user geographical location, user consumption frequency, user consumption level, and user interaction times corresponding to the user profile, where the target push product represents a business product for which the user's purchase times are greater than a first preset threshold, and the user interaction times represent the number of times the user participates in historical push activities; Step S2032: Determine a push mechanism according to the target push product, the user geographical location, the user consumption frequency, the user consumption level, and the user interaction times, and push business products to the user according to the push mechanism. Through the above steps, this method converts the user profile into a specific and effective push strategy. Through the refined design of the push mechanism, financial institutions can improve the utilization rate and conversion rate of the preferential mechanism, reduce operating costs, and improve operating efficiency.

[0052] In the specific implementation process, the historical purchase records and consumption preferences in the user profile can help financial institutions identify which business products are of high interest to users. If the purchase frequency of a certain product exceeds a preset first threshold, this product becomes a target push product. For example, if a user frequently uses a credit card of a certain financial institution to consume in travel or dining scenarios, then credit card promotion activities or products related to travel or dining will be defined as target push products. The geographical location information in the user profile can be used to identify the user's permanent residence or activity range, which is particularly important for the promotion of preferential mechanisms in offline stores. For example, if a user often activities in a specific area, the financial institution can give priority to pushing preferential information within this area to increase the likelihood of offline consumption. The user's consumption frequency and consumption level are key indicators for evaluating the user's consumption ability and habits. For users with high consumption frequency and high consumption level, high-end products or high-value preferential mechanisms can be pushed to meet their consumption needs and strengthen their loyalty. For users with low consumption frequency and average consumption level, more user-friendly promotion activities can be designed to stimulate their consumption interest. The user profile also includes the interaction records between the user and the financial institution, including the number of times of participating in historical push activities. Users with more interaction times may be more inclined to participate in the activities of the financial institution. The financial institution can provide more participation opportunities for such users or design promotion activities with strong interactivity to enhance their sense of participation and satisfaction. Based on the above analysis results, a personalized push mechanism can be determined. The push mechanism should take into account the customization of push time, push channels, and push content to improve the push efficiency and user response rate. For example, for highly active users, it can be pushed in the afternoon on weekdays through the mobile financial institution APP or by text message; for users with obvious consumption preferences, the push content should directly target their preferred products or services. Finally, according to the determined push mechanism, business products or preferential information are pushed to users through corresponding channels. After the push, the financial institution needs to track the feedback and behavior of users, collect data for subsequent analysis and model optimization to continuously improve the push effect.

[0053] In some alternative embodiments, step S2032 determines a push mechanism based on the target push product, the user's geographical location, the user's consumption frequency, the user's consumption level, and the number of user interactions, and can be implemented through the following steps: Push the target push product to the user again; obtain the business product address, calculate the absolute value of the difference in the distance between the business product address and the user's geographical location, and push the business product corresponding to the business product address whose absolute value of the difference is less than a second preset threshold to the user; when the user's consumption frequency is greater than a third preset threshold or the user's consumption level is greater than a fourth preset threshold or the number of user interactions is greater than a fifth preset threshold, push a preferential mechanism to the user. Through the above steps, this method can not only customize the push plan according to the static information of the user portrait (such as consumption preferences, static tags), but also combine the user's dynamic behaviors (geographical location, consumption frequency, consumption level, number of interactions) to adjust the preferential mechanism and push content in real time, so as to achieve more accurate and personalized push.

[0054] Specifically, pushing the target product to the user again: If the previous push activity did not obtain the expected response, or the financial institution hopes to strengthen the user's interest in a specific product, the target product can be pushed to the user again based on the user portrait. This usually occurs when the user portrait shows that the user has continuous or potential interest in the product, such as the user has a past purchase record, or shows attention to this type of product in behaviors such as browsing and searching. Calculating the proximity of the product address to the user's geographical location: The financial institution or business platform has detailed address information of the business product. By calculating the absolute value of the difference in the distance between the product address and the user's permanent residence or current geographical location, it can be identified which products are close to the user's geographical location. If this absolute value is less than the second preset threshold, then these geographically close products will be preferentially pushed. This helps to attract users to the store for consumption, especially when the user has an immediate need or geographical factors have a greater impact on the consumption decision. Pushing the preferential mechanism according to the user's consumption frequency and consumption level: If the user's consumption frequency is higher than the third preset threshold, it indicates that the user is a frequent consumer, and the financial institution can push preferential mechanisms such as higher cashback, points rewards, or membership upgrades to encourage continuous consumption. Similarly, if the user's consumption level is higher than the fourth preset threshold, it indicates that the user has strong consumption ability, and the financial institution can try to provide higher-quality products or larger discounts for the user. Preferential push based on the number of user interactions: The number of interactions in the user portrait reflects the enthusiasm of the user to participate in the activities of the financial institution. If the number of user interactions exceeds the fifth preset threshold, the financial institution can push preferential activities with lower participation thresholds and strong interactivity, such as lucky draws, voting, or feedback surveys, to further enhance user participation and brand loyalty.

[0055] In some alternative embodiments, before analyzing the user characteristics through the user profile model, the method further includes the following steps: constructing an initial user profile model and obtaining the historical user characteristics, where the initial user profile model includes an input layer, a hidden layer, and an output layer; inputting the historical user characteristics into the input layer of the initial user profile model, and analyzing the historical user characteristics through the hidden layer until the loss function is minimized to obtain the user profile model. Through the above steps, the method constructs a deep learning model based on historical user characteristics, which can extract meaningful feature representations from complex user data for generating a more accurate and comprehensive user profile.

[0056] Specifically, the design of the initial user profile model is usually based on the multi-layer perceptron (MLP) architecture, including an input layer, one or more hidden layers, and an output layer. The input layer receives historical user characteristics, the hidden layer is responsible for performing complex non-linear transformations on the input features to extract deep feature representations, and the output layer gives the predicted output of the user characteristics. Structurally, the MLP extends a single-layer network of a linear model to multiple layers. The output of each layer is used as the input of the next layer's feed-forward network after passing through a non-linear transformation, thus obtaining a forward network with information transmitted layer by layer. The input layer receives the vector of historical user characteristics as input; the hidden layer can include multiple fully connected layers, and each layer introduces non-linearity through an activation function; the output layer outputs the predicted probability of the user for the target category.

[0057] In some alternative embodiments, analyzing the historical user characteristics through the hidden layer can be achieved through the following steps: analyzing the historical user characteristics through the function h of the hidden layer (l) = σ(W (l) h (l-1) + b (l) ), where σ represents the activation function, W (l) represents the weight of the l-th layer, b (l) represents the bias of the l-th layer, h (l-1) represents the output result of the hidden layer of the (l - 1)-th layer, and h (l) represents the output result of the hidden layer of the l-th layer. Through the hidden layer of the multi-layer perceptron, this method can extract deeper information from historical user characteristics, providing strong support for the construction of the user profile.

[0058] In the specific implementation process, the output of the hidden layer is the above h (l) = σ(W (l) h (l-1) + b (l) ); in addition, the predicted result of the user category predicted by the output layer is y pred = softmax(W (L) h(L-1) +b (L) ), where L is the index of the output layer, and the softmax function converts the output into a probability distribution. Then, model training and optimization are carried out: define the loss function, and use the cross-entropy loss function where m is the number of samples and k is the number of classes. Calculate the gradient using the backpropagation algorithm and update the model parameters using an optimization algorithm to minimize the loss function. Add L2 regularization to prevent overfitting. Finally, divide the training set and the validation set, and evaluate the model performance using the validation set. Adjust hyperparameters such as the learning rate and the number of nodes in the hidden layer to optimize the model performance. Use 70% of the obtained data for training and 30% for validation. After training, predict the features of new users. Substitute the features of new users into the trained model for prediction to obtain the prediction probabilities of users for different classes. Generate user portraits based on the prediction results.

[0059] To enable those skilled in the art to more clearly understand the technical solution of this application, the implementation process of the financial business push method of this application will be described in detail below with specific embodiments.

[0060] This embodiment relates to a specific financial business push method, such as Figure 3 shown, including the following steps:

[0061] Step S1: Apply the user portrait result to the financial business platform;

[0062] Step S2: Push the preferential mechanism and handle the business;

[0063] Step S3: Develop a preferential mechanism according to consumption preferences: According to the user's consumption preferences and historical purchase records, issue a preferential mechanism for the products or categories they like to attract them to purchase again;

[0064] Step S4: Set the preferential mechanism based on the consumption frequency: For high-frequency consumption customers, preferential mechanisms such as cashback or doubled points can be set to encourage them to continue consuming;

[0065] Step S5: Provide preferential mechanisms according to the geographical location: Combine the user's geographical location information and send preferential mechanisms to customers near the store to increase the possibility of in-store consumption;

[0066] Step S6: Set up a stepped discount according to the consumption level: Provide high-value preferential mechanisms or discounts for high-consumption customers, and provide appropriate promotional activities for low-consumption customers, and customize the preferential according to the consumption level;

[0067] Step S7: Design a preferential mechanism in combination with the user's interaction behavior: Provide exclusive preferential mechanisms for customers who participate in more activities or have positive interaction behaviors to enhance user participation and loyalty;

[0068] Step S8: Based on the user portrait results, conduct business preferential applications and identify abnormal situations for customers, such as preferential fee collection, abnormal fund behavior, or fraud risks during the business handling process, so as to effectively protect the interests of financial institutions and customers.

[0069] The embodiment of the present application also provides a financial business push device. It should be noted that the Z device in the embodiment of the present application can be used to execute the financial business push method provided by the embodiment of the present application. This device is used to implement the above embodiments and preferred implementation manners, and those that have been described will not be repeated. As used hereinafter, the term "module" can be a combination of software and / or hardware that realizes a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware is also possible and contemplated.

[0070] The following introduces the financial business push device provided by the embodiment of the present application.

[0071] Figure 4 It is a schematic diagram of the financial business push device according to the embodiment of the present application. As Figure 4 shown, the device includes:

[0072] An extraction unit 10, configured to obtain user data and extract user features according to the user data, where the user data includes user identity information and user consumption information, and the user features characterize the consumption characteristics of the user;

[0073] Specifically, through user authorization, collect the user's personal information data and consumption habit data. Obtaining user data and extracting user features is the first and crucial step in building a user portrait. User data can be collected from multiple channels, including but not limited to: User identity information: including basic information such as the user's age, gender, occupation, educational background, and marital status, which helps to understand the user's background characteristics and thus infer their possible consumption preferences and behavior patterns. User consumption information: covers the user's consumption records, including consumption time, consumption location, consumption amount, consumption category, payment method, etc., which reflect the user's actual consumption behavior and consumption habits. The user features extracted from these data can include: consumption preferences, consumption frequency, consumption amount, geographical location, and user engagement. When extracting user features, the following technical means can be adopted: data cleaning, feature engineering, and then through a machine learning model: using deep learning models such as a multi-layer perceptron (MLP) or an RFM model, etc., to analyze the user data, extract key features, and build a user portrait.

[0074] The first analysis unit 20 is configured to analyze the user features through a user portrait model to obtain a user category prediction result, where the user category prediction result represents the consumption level of the user. The user portrait model is trained with multiple sets of data, and each set of data in the multiple sets of data includes historical user features and the user category prediction result corresponding to the historical user features;

[0075] Specifically, the process of analyzing user features through a user portrait model to obtain a user category prediction result is an advanced data analysis based on deep learning techniques (such as multi-layer perceptron, MLP). This process can be divided into several key steps: Data collection and preprocessing: First, a large amount of user data needs to be collected, including user identity information, consumption information, etc., to form historical user features. Preprocessing includes data cleaning, missing value handling, feature selection, feature encoding, etc., to ensure data quality and model training effect. Feature and label preparation: For each set of historical data, extract user features (such as age, gender, consumption amount, consumption frequency, etc.) and label the consumption level of the user (such as high-level, mid-level, low-level consumers). The label of the consumption level is usually obtained through a comprehensive analysis of the customer's consumption behavior and consumption amount. Model training: Use a multi-layer perceptron model for training. A multi-layer perceptron is a neural network for supervised learning that can learn the complex relationship between input features and output labels. In the training stage, the model predicts the consumption level based on the input user features, and continuously adjusts the weights and biases of the model through a loss function (such as cross-entropy loss) and backpropagation algorithm to make the prediction result as close as possible to the actual label.

[0076] In this process, the key lies in training a model through a large amount of historical data that can accurately predict the user's consumption level, so as to help financial institutions better understand and predict user behavior, and formulate more effective preferential strategies. In this way, financial institutions can improve the push efficiency, reduce ineffective delivery, and enhance customer satisfaction and loyalty. At the same time, based on the prediction and classification of user behavior, financial institutions can also develop more customized products and services to meet the needs of customers at different consumption levels, thereby improving the market competitiveness and customer service level of financial institutions.

[0077] The push unit 30 is configured to generate a user portrait according to the user category prediction result, and perform business push on the user according to the user portrait, where the user portrait at least includes the consumption features of the user.

[0078] Specifically, after analyzing user characteristics through a user portrait model (such as an MLP model), user category prediction results are obtained. These prediction results can be combinations of labels such as consumption level, consumption preferences, and consumption frequency, thereby constructing a complete user portrait. The generation of the user portrait is not just a simple aggregation of prediction results. It also requires combining the prediction results with other information of the user (such as basic information, geographical location, hobbies, etc.) to form a more comprehensive and in-depth understanding of the user. The user portrait can include, but is not limited to, the following information: Consumption characteristics: including consumption preferences, consumption frequency, consumption amount, etc. User basic information: age, gender, occupation, educational background, marital status, etc. Geographical location: frequently visited locations, place of residence, place of work, etc. User behavior characteristics: frequency of participating in activities, feedback, purchase channel preferences, etc. With the user portrait, financial institutions can formulate more precise business push strategies based on the consumption characteristics and preferences of the users in the portrait.

[0079] In this embodiment, user data is obtained, and user characteristics are extracted according to the user data. Among them, the user data includes user identity information and user consumption information, and the user characteristics represent the consumption characteristics of the user. The user portrait model analyzes the user characteristics to obtain user category prediction results. Among them, the user category prediction results represent the consumption level of the user. The user portrait model is trained through multiple sets of data, and each set of data in the multiple sets of data includes historical user characteristics and the user category prediction results corresponding to the historical user characteristics. A user portrait is generated according to the user category prediction results, and business push is performed on the user according to the user portrait. Among them, the user portrait at least includes the consumption characteristics of the user. Compared with the prior art where financial services cannot be accurately pushed, in this application, user characteristics are extracted to generate a user portrait, and business push is performed according to the user portrait, so that accurate push can be performed according to the consumption characteristics of the user, improving the user experience. Therefore, the problem of inaccurate push in the prior art can be solved, and the effect of accurately pushing financial services to relevant users can be achieved.

[0080] In the specific implementation process, before extracting user characteristics according to the user data, the device further includes a labeling unit and a standardization processing unit. The labeling unit is used to label the user data to obtain static labels and dynamic labels. Among them, the static labels represent the unchanging user data, and the dynamic labels represent the user data that changes dynamically. The standardization processing unit is used to perform standardization processing on the user data labeled with the static labels and the dynamic labels to obtain the standardized user data. The device performs standardization processing, and the standardized data helps the model better identify and learn the consumption characteristics of the user, avoiding model skew caused by differences in data dimensions or ranges. The real-time update of the dynamic labels ensures the timeliness and accuracy of the user portrait, enabling financial institutions to adjust preferential policies in a timely manner to adapt to changes in user behavior.

[0081] Specifically, data annotation is the process of converting raw data into a machine-readable and processable format. At this stage, the collected user data is analyzed to distinguish which data is static and which data is dynamically changing. Static labels: Such labels represent the user's unchanging basic information, such as age, gender, occupation, educational background, marital status, etc. These information usually do not change frequently and are the basis for constructing user portraits. Dynamic labels: In contrast to static labels, dynamic labels record the user's behavioral information that changes over time, such as recent consumption records, activity participation, interaction feedback, etc. Dynamic labels are crucial for capturing the user's current preferences and behavior patterns. Data standardization processing: Data standardization processing is the process of converting user data to the same scale, aiming to eliminate the differences in dimension and numerical range between different features, so that the model can learn more effectively. The standardization processing usually includes the following steps: Numerical conversion: Convert non-numerical data (such as gender, occupation) into numerical data for easy model processing. Common conversion methods include one-hot encoding, label encoding, etc. Data normalization: Scale numerical data to a fixed range, usually the [0, 1] interval, to avoid the impact of numerical differences between features on model training. Normalization methods include min-max scaling, z-score standardization, etc. Missing value processing: Check and process missing values in the data, and devices such as deletion, filling with the mean, and predictive filling can be used to ensure the integrity of the data and the accuracy of model training.

[0082] In some alternative embodiments, the extraction unit includes a first determination module and a second determination module. The first determination module is used to determine the value of the user data through a value evaluation model; the second determination module is used to obtain the one-to-one mapping relationship between the value and the preset user characteristics, and determine the preset user characteristics corresponding to the value of the user data according to the one-to-one mapping relationship, so as to obtain the user characteristics. Through the above steps, the device can convert the value of the user data into an in-depth understanding of the user characteristics, and then provide more accurate and personalized services and discounts for different user groups. At the same time, it can effectively identify and control risks, and improve the overall business operation efficiency and customer satisfaction.

[0083] In the specific implementation process, a value evaluation model needs to be established first. This model can quantify the value of user data to the business. The value evaluation model may be based on multiple factors, such as the consumption ability, loyalty, activity level, and potential risks of users. Value evaluation can be carried out through the RFM model. The RFM model is a customer classification model used to describe the customer value situation and is a data analysis model for measuring customer value and customer profit-making ability. It describes the customer value situation through three key dimensions: Recency (the time of the last consumption), Frequency (the consumption frequency), and Monetary (the consumption amount). The preset user characteristics refer to a series of tags or indicators predefined by financial institutions to describe user attributes and behaviors. Before establishing the mapping relationship with the value, the financial institution needs to define which user characteristics correspond to high-value, medium-value, or low-value user data. For example, users with high consumption frequency, high consumption amount, and active participation in financial institution activities may be defined as high-value, while users with low consumption and less participation may be defined as low-value. Once the value evaluation model gives the value of the user data, the financial institution can determine the user characteristics corresponding to the value of the user data based on the preset value characteristic mapping relationship. For example, if a certain user data is evaluated as high-value, the financial institution may determine that the user has characteristics such as high consumption ability, high usage frequency, or high loyalty. According to the obtained user characteristics, the financial institution can customize service strategies, such as launching high-end product or service discounts for users with high consumption ability, and increasing integral rewards or membership upgrade opportunities for users with high activity levels. In addition, the financial institution can also identify potential business risks, such as fraud risks and credit risks, through user characteristics, so as to manage risks in a more efficient way.

[0084] In some alternative implementation manners, the above-mentioned push unit includes an acquisition module and a determination module. The acquisition module is used to acquire the target push product corresponding to the user portrait, the user geographical location, the user consumption frequency, the user consumption level, and the user interaction times. Among them, the target push product represents a business product with the number of purchases by the user greater than a first preset threshold, and the user interaction times represent the number of times the user participates in historical push activities; the determination module is used to determine a push mechanism according to the target push product, the user geographical location, the user consumption frequency, the user consumption level, and the user interaction times, and push the business product to the user according to the push mechanism. Through the above steps, the device transforms the user portrait into a specific and effective push strategy. Through the refined design of the push mechanism, the financial institution can improve the usage rate and conversion rate of the preferential mechanism, reduce the operation cost, and improve the operation efficiency.

[0085] In the specific implementation process, the historical purchase records and consumption preferences in the user profile can help financial institutions identify which business products are of high interest to users. If the purchase frequency of a certain product exceeds a preset first threshold, this product becomes a target push product. For example, if a user frequently uses a credit card of a certain financial institution for consumption in travel or dining scenarios, then credit card promotion activities or products related to travel or dining will be defined as target push products. The geographical location information in the user profile can be used to identify the user's permanent residence or activity range, which is particularly important for the promotion of preferential mechanisms in offline stores. For example, if a user often activities in a specific area, the financial institution can give priority to pushing preferential information within this area to increase the likelihood of offline consumption. The user's consumption frequency and consumption level are key indicators for evaluating the user's consumption ability and habits. For users with high consumption frequency and high consumption level, high-end products or high-value preferential mechanisms can be pushed to meet their consumption needs and strengthen their loyalty. For users with low consumption frequency and average consumption level, more user-friendly promotion activities can be designed to stimulate their consumption interest. The user profile also includes the interaction records between the user and the financial institution, including the number of times of participating in historical push activities. Users with a large number of interactions may be more inclined to participate in the activities of the financial institution. The financial institution can provide more participation opportunities for such users or design promotion activities with strong interactivity to enhance their sense of participation and satisfaction. Based on the above analysis results, a personalized push mechanism can be determined. The push mechanism should consider the customization of push time, push channels, and push content to improve the push efficiency and user response rate. For example, for highly active users, it can be pushed in the afternoon on weekdays through the mobile financial institution APP or SMS; for users with obvious consumption preferences, the push content should be directly targeted at their preferred products or services. Finally, according to the determined push mechanism, business products or preferential information are pushed to users through corresponding channels. After the push, the financial institution needs to track the feedback and behavior of users, collect data for subsequent analysis and model optimization to continuously improve the push effect.

[0086] In some alternative embodiments, the determination module includes a first push sub-module, a second push sub-module, and a third push sub-module. The first push sub-module is configured to push the target push product to the user again; the second push sub-module is configured to obtain the business product address and calculate the absolute value of the difference between the business product address and the user's geographical location, and push the business product corresponding to the business product address whose absolute value of the difference is less than a second preset threshold to the user; the third push sub-module is configured to push a preferential mechanism to the user when the user's consumption frequency is greater than a third preset threshold or the user's consumption level is greater than a fourth preset threshold or the user's interaction times is greater than a fifth preset threshold. Through the above steps, the device can not only customize the push scheme according to the static information of the user portrait (such as consumption preferences, static tags), but also combine the user's dynamic behaviors (geographical location, consumption frequency, consumption level, interaction times) to adjust the preferential mechanism and push content in real time, so as to achieve more accurate and personalized push.

[0087] Specifically, pushing the target product to the user again: If the previous push activity did not obtain the expected response, or the financial institution hopes to strengthen the user's interest in a specific product, the target product can be pushed to the user again based on the user portrait. This usually occurs when the user portrait shows that the user has continuous or potential interest in the product, such as the user has a past purchase record, or shows concern for this type of product in behaviors such as browsing and searching. Calculating the proximity of the product address to the user's geographical location: The financial institution or business platform has detailed address information of the business product. By calculating the absolute value of the difference between the product address and the user's permanent location or current geographical location, it can be identified which products are close to the user's geographical location. If this absolute value is less than the second preset threshold, then these geographically close products will be preferentially pushed. This helps to attract users to the store for consumption, especially when the user has an immediate need or geographical factors have a greater impact on the consumption decision. Pushing the preferential mechanism according to the user's consumption frequency and consumption level: If the user's consumption frequency is higher than the third preset threshold, indicating that the user is a frequent consumer, the financial institution can push preferential mechanisms such as higher cashback, point rewards, or membership upgrades to encourage continuous consumption. Similarly, if the user's consumption level is higher than the fourth preset threshold, indicating that the user has a strong consumption ability, the financial institution can try to provide higher-quality products or larger discounts for the user. Preferential push based on the user's interaction times: The interaction times in the user portrait reflect the user's enthusiasm for participating in the activities of the financial institution. If the user's interaction times exceed the fifth preset threshold, the financial institution can push preferential activities with lower participation thresholds and stronger interactivity to the user, such as lucky draws, voting, or feedback surveys, to further enhance the user's participation and brand loyalty.

[0088] In some alternative embodiments, before analyzing the user features through the user portrait model, the device further includes an acquisition unit and a second analysis unit. The acquisition unit is configured to construct an initial user portrait model and acquire the historical user features, where the initial user portrait model includes an input layer, a hidden layer, and an output layer. The second analysis unit is configured to input the historical user features into the input layer of the initial user portrait model and analyze the historical user features through the hidden layer until the loss function is minimized to obtain the user portrait model. The device constructs a deep learning model based on historical user features through the above steps. This model can extract meaningful feature representations from complex user data for generating a more accurate and comprehensive user portrait.

[0089] Specifically, the design of the initial user portrait model is usually based on the multi-layer perceptron (MLP) architecture, including an input layer, one or more hidden layers, and an output layer. The input layer receives historical user features, and the hidden layer is responsible for performing complex non-linear transformations on the input features to extract deep feature representations. The output layer then gives the predicted output of the user features. Structurally, the MLP extends a single-layer network of a linear model to multiple layers. The output of each layer is used as the input of the next layer's feed-forward network after passing through a non-linear transformation, thereby obtaining a forward network with information propagating layer by layer. The input layer receives a vector of historical user features as input. The hidden layer can include multiple fully connected layers, and each layer introduces non-linearity through an activation function. The output layer outputs the predicted probability of the user for the target category.

[0090] In some alternative embodiments, the second analysis unit includes an analysis module for analyzing the historical user features through the function h of the hidden layer (l) =σ(W (l) h (l-1) +b (l) ), where σ represents the activation function, W (l) represents the weight of the l-th layer, b (l) represents the bias of the l-th layer, h (l-1) represents the output result of the hidden layer of the (l - 1)-th layer, and h (l) represents the output result of the hidden layer of the l-th layer. Through the hidden layer of the multi-layer perceptron, the device can extract deeper information from historical user features, providing strong support for the construction of the user portrait.

[0091] In the specific implementation process, the output of the hidden layer is the above h (l) =σ(W (l) h (l-1) +b (l) ); in addition, the predicted result of the user category predicted by the output layer is y pred =softmax(W (L) h(L-1) +b (L) ), where L is the index of the output layer, and the softmax function converts the output into a probability distribution. Then, model training and optimization are carried out: Define the loss function, and use the cross-entropy loss function where m is the number of samples and k is the number of classes. Calculate the gradient using the backpropagation algorithm and update the model parameters using the optimization algorithm to minimize the loss function. Add L2 regularization to prevent overfitting. Finally, divide the training set and the validation set, and evaluate the model performance using the validation set. Adjust hyperparameters such as the learning rate and the number of hidden layer nodes to optimize the model performance. Use 70% of the obtained data for training and 30% for validation. After training, predict the characteristics of new users. Substitute the characteristics of new users into the trained model for prediction to obtain the prediction probabilities of users for different classes. Generate user portraits according to the prediction results.

[0092] The push device for financial services includes a processor and a memory. The above extraction unit, first analysis unit, push unit, etc. are all stored in the memory as program units, and the corresponding functions are implemented by the processor executing the above program units stored in the memory. The above modules are all located in the same processor; or, the above modules are respectively located in different processors in any combination form.

[0093] The processor contains a kernel, and the kernel retrieves the corresponding program unit from the memory. One or more kernels can be set, and by adjusting the kernel parameters, the problem that the financial services in the prior art cannot be accurately pushed, resulting in poor user experience, can be solved.

[0094] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash memory (flash RAM), and the memory includes at least one storage chip.

[0095] An embodiment of the present invention provides a computer-readable storage medium, and the computer-readable storage medium includes a push method for financial services.

[0096] Specifically, the push method for financial services includes:

[0097] Step S201, obtain user data, extract user characteristics according to the user data, where the user data includes user identity information and user consumption information, and the user characteristics represent the consumption characteristics of users;

[0098] Specifically, through user authorization, personal information data and consumption habit data of users are collected. Obtaining user data and extracting user characteristics is the first and crucial step in building a user portrait. User data can be collected from multiple channels, including but not limited to: User identity information: including basic information such as the user's age, gender, occupation, educational background, marital status, etc. These information help to understand the background characteristics of the user, thereby inferring their possible consumption preferences and behavior patterns. User consumption information: covering the user's consumption records, including consumption time, consumption location, consumption amount, consumption category, payment method, etc. These data reflect the user's actual consumption behavior and consumption habits. User characteristics extracted from these data can include: consumption preferences, consumption frequency, consumption amount, geographical location, user engagement. When extracting user characteristics, the following technical means can be adopted: data cleaning, feature engineering, and then through machine learning models: using deep learning models such as multi-layer perceptron (MLP) or RFM models, etc., to analyze user data, extract key features, and build a user portrait.

[0099] Step S202, analyze the user characteristics through the user portrait model to obtain a user category prediction result, where the user category prediction result represents the consumption level of the user, and the user portrait model is trained through multiple sets of data, and each set of data in the multiple sets of data includes historical user characteristics and the user category prediction result corresponding to the historical user characteristics;

[0100] Specifically, the process of analyzing user characteristics through the user portrait model to obtain a user category prediction result is an advanced data analysis based on deep learning technology (such as multi-layer perceptron, MLP). This process can be divided into several key steps: Data collection and preprocessing: First, a large amount of user data needs to be collected, including user identity information, consumption information, etc., to form historical user characteristics. Preprocessing includes data cleaning, missing value processing, feature selection, feature encoding, etc., to ensure data quality and model training effect. Feature and label preparation: For each set of historical data, extract user characteristics (such as age, gender, consumption amount, consumption frequency, etc.) and label the consumption level of the user (such as high-level, middle-level, low-level consumers). The label of the consumption level is usually obtained based on the comprehensive analysis of the customer's consumption behavior and consumption amount. Model training: Use a multi-layer perceptron model for training. The multi-layer perceptron is a neural network for supervised learning that can learn the complex relationship between input features and output labels. In the training stage, the model will predict the consumption level based on the input user characteristics, and continuously adjust the weights and biases of the model through a loss function (such as cross-entropy loss) and backpropagation algorithm to make the prediction result as close as possible to the actual label.

[0101] In this process, the key lies in training a model through a large amount of historical data that can accurately predict the user's consumption level, so as to help financial institutions better understand and predict user behavior, and formulate more effective preferential strategies. In this way, financial institutions can improve the push efficiency, reduce ineffective delivery, and enhance customer satisfaction and loyalty. At the same time, based on the prediction and classification of user behavior, financial institutions can also develop more customized products and services to meet the needs of customers at different consumption levels, thereby improving the market competitiveness and customer service level of financial institutions.

[0102] Step S203, generate a user portrait according to the user category prediction result, and perform business push on the user according to the user portrait, where the user portrait at least includes the consumption characteristics of the user.

[0103] Specifically, when analyzing user characteristics through a user portrait model (such as an MLP model), user category prediction results are obtained. These prediction results can be a combination of labels such as consumption level, consumption preference, and consumption frequency, so as to construct a complete user portrait. The generation of a user portrait is not just a simple aggregation of prediction results, but also needs to combine the prediction results with other information of the user (such as basic information, geographical location, hobbies, etc.) to form a more comprehensive and in-depth understanding of the user. The user portrait may include, but is not limited to, the following information: Consumption characteristics: including consumption preferences, consumption frequency, consumption amount, etc. User basic information: age, gender, occupation, educational background, marital status, etc. Geographical location: frequently visited places, place of residence, place of work, etc. User behavior characteristics: frequency of participating in activities, feedback, purchase channel preferences, etc. With the user portrait, financial institutions can formulate more accurate business push strategies based on the consumption characteristics and preferences of the users in the portrait.

[0104] An embodiment of the present invention provides an electronic device, including a processor, a memory, and a program stored on the memory and executable on the processor. When the processor executes the program, it implements at least the following steps:

[0105] Step S201, obtain user data, and extract user characteristics according to the user data, where the user data includes user identity information and user consumption information, and the user characteristics characterize the consumption characteristics of the user;

[0106] Specifically, through user authorization, personal information data and consumption habit data of users are collected. Obtaining user data and extracting user features is the first and crucial step in constructing a user portrait. User data can be collected from multiple channels, including but not limited to: User identity information: including basic information such as the user's age, gender, occupation, educational background, marital status, etc. These information help to understand the background characteristics of the user, thereby inferring their possible consumption preferences and behavior patterns. User consumption information: covering the user's consumption records, including consumption time, consumption location, consumption amount, consumption categories, payment methods, etc. These data reflect the user's actual consumption behavior and consumption habits. User features extracted from these data can include: consumption preferences, consumption frequency, consumption amount, geographical location, user engagement. When extracting user features, the following technical means can be adopted: data cleaning, feature engineering, and then through machine learning models: using deep learning models such as multi-layer perceptron (MLP) or RFM models, etc., to analyze user data, extract key features, and build a user portrait.

[0107] Step S202, analyze the user features through the user portrait model to obtain a user category prediction result, where the user category prediction result represents the consumption level of the user, and the user portrait model is trained through multiple sets of data, and each set of data in the multiple sets of data includes historical user features and the user category prediction result corresponding to the historical user features;

[0108] Specifically, the process of analyzing user features through the user portrait model to obtain a user category prediction result is an advanced data analysis based on deep learning technology (such as multi-layer perceptron, MLP). This process can be divided into several key steps: Data collection and preprocessing: First, a large amount of user data needs to be collected, including user identity information, consumption information, etc., to form historical user features. Preprocessing includes data cleaning, missing value processing, feature selection, feature encoding, etc., to ensure data quality and model training effect. Feature and label preparation: For each set of historical data, extract user features (such as age, gender, consumption amount, consumption frequency, etc.) and label the consumption level of the user (such as high-level, mid-level, low-level consumers). The label of the consumption level is usually obtained based on the comprehensive analysis of the user's consumption behavior and consumption amount. Model training: Use the multi-layer perceptron model for training. The multi-layer perceptron is a neural network for supervised learning that can learn the complex relationship between input features and output labels. In the training stage, the model will predict the consumption level based on the input user features, and continuously adjust the weights and biases of the model through a loss function (such as cross-entropy loss) and backpropagation algorithm to make the prediction result as close as possible to the actual label.

[0109] In this process, the key lies in training a model through a large amount of historical data that can accurately predict the user's consumption level, thereby helping financial institutions better understand and predict user behavior, and formulate more effective preferential strategies. In this way, financial institutions can improve the push efficiency, reduce ineffective deliveries, and enhance customer satisfaction and loyalty. At the same time, based on the prediction and classification of user behavior, financial institutions can also develop more customized products and services to meet the needs of customers at different consumption levels, thereby improving the market competitiveness and customer service level of financial institutions.

[0110] Step S203: Generate a user portrait according to the user category prediction result, and perform business push on the user according to the user portrait, where the user portrait at least includes the consumption characteristics of the user.

[0111] Specifically, when analyzing user characteristics through a user portrait model (such as an MLP model), user category prediction results are obtained. These prediction results can be a combination of labels such as consumption level, consumption preference, and consumption frequency, so as to construct a complete user portrait. The generation of the user portrait is not just a simple aggregation of prediction results, but also needs to combine the prediction results with other information of the user (such as basic information, geographical location, hobbies, etc.) to form a more comprehensive and in-depth understanding of the user. The user portrait can include, but is not limited to, the following information: Consumption characteristics: including consumption preferences, consumption frequency, consumption amount, etc. User basic information: age, gender, occupation, educational background, marital status, etc. Geographical location: frequently visited locations, place of residence, place of work, etc. User behavior characteristics: frequency of participating in activities, feedback, purchase channel preferences, etc. With the user portrait, financial institutions can formulate more accurate business push strategies based on the consumption characteristics and preferences of the users in the portrait.

[0112] The devices in this article can be servers, PCs, PADs, mobile phones, etc.

[0113] This application also provides a computer program product, including a computer program, which when executed by a processor implements the steps of the methods in the various embodiments of this application: Step S201: Obtain user data, and extract user characteristics according to the user data, where the user data includes user identity information and user consumption information, and the user characteristics characterize the consumption characteristics of the user;

[0114] Specifically, through user authorization, personal information data and consumption habit data of users are collected. Obtaining user data and extracting user features is the first and crucial step in building a user portrait. User data can be collected from multiple channels, including but not limited to: User identity information: including basic information such as the user's age, gender, occupation, educational background, marital status, etc. These information help to understand the background characteristics of the user, so as to infer their possible consumption preferences and behavior patterns. User consumption information: covering the user's consumption records, including consumption time, consumption location, consumption amount, consumption category, payment method, etc. These data reflect the user's actual consumption behavior and consumption habits. The user features extracted from these data can include: consumption preferences, consumption frequency, consumption amount, geographical location, user engagement. When extracting user features, the following technical means can be adopted: data cleaning, feature engineering, and then through machine learning models: using deep learning models such as multi-layer perceptron (MLP) or RFM models, etc., to analyze user data, extract key features, and build a user portrait.

[0115] Step S202, analyze the user features through the user portrait model to obtain a user category prediction result, where the user category prediction result represents the consumption level of the user, the user portrait model is trained through multiple groups of data, and each group of data in the multiple groups of data includes historical user features and the user category prediction result corresponding to the historical user features;

[0116] Specifically, the process of analyzing user features through the user portrait model to obtain a user category prediction result is an advanced data analysis based on deep learning technology (such as multi-layer perceptron, MLP). This process can be divided into several key steps: Data collection and preprocessing: First, a large amount of user data needs to be collected, including user identity information, consumption information, etc., to form historical user features. Preprocessing includes data cleaning, missing value processing, feature selection, feature encoding, etc., to ensure data quality and model training effect. Feature and label preparation: For each group of historical data, extract user features (such as age, gender, consumption amount, consumption frequency, etc.) and label the consumption level of the user (such as high-level, middle-level, low-level consumers). The label of the consumption level is usually obtained based on the comprehensive analysis of the user's consumption behavior and consumption amount. Model training: Use a multi-layer perceptron model for training. The multi-layer perceptron is a neural network for supervised learning that can learn the complex relationship between input features and output labels. In the training stage, the model will predict the consumption level based on the input user features, and continuously adjust the weights and biases of the model through a loss function (such as cross-entropy loss) and backpropagation algorithm, so that the prediction result is as close as possible to the actual label.

[0117] In this process, the key lies in training a model through a large amount of historical data that can accurately predict the user's consumption level, so as to help financial institutions better understand and predict user behavior, and formulate more effective preferential strategies. In this way, financial institutions can improve the push efficiency, reduce ineffective deliveries, and enhance customer satisfaction and loyalty. At the same time, based on the prediction and classification of user behavior, financial institutions can also develop more customized products and services to meet the needs of customers at different consumption levels, thereby improving the market competitiveness and customer service level of financial institutions.

[0118] Step S203: Generate a user portrait according to the user category prediction result, and perform business push on the user according to the user portrait, where the user portrait at least includes the consumption characteristics of the user.

[0119] Specifically, after analyzing user characteristics through a user portrait model (such as an MLP model), user category prediction results are obtained. These prediction results can be a combination of labels such as consumption level, consumption preference, and consumption frequency, so as to construct a complete user portrait. The generation of a user portrait is not just a simple aggregation of prediction results, but also needs to combine the prediction results with other information of the user (such as basic information, geographical location, hobbies, etc.) to form a more comprehensive and in-depth understanding of the user. The user portrait may include, but is not limited to, the following information: Consumption characteristics: including consumption preferences, consumption frequency, consumption amount, etc. User basic information: age, gender, occupation, educational background, marital status, etc. Geographical location: frequently visited places, place of residence, place of work, etc. User behavior characteristics: frequency of participating in activities, feedback, preference for purchase channels, etc. With the user portrait, financial institutions can formulate more accurate business push strategies based on the consumption characteristics and preferences of the users in the portrait.

[0120] Obviously, those skilled in the art should understand that the above-mentioned modules or steps of the present invention can be implemented by a general-purpose computing device. They can be concentrated on a single computing device or distributed on a network composed of multiple computing devices. They can be implemented by program codes executable by the computing device, so that they can be stored in a storage device and executed by the computing device. And in some cases, the steps shown or described can be executed in a different order from here, or they can be separately made into individual integrated circuit modules, or multiple modules or steps among them can be made into a single integrated circuit module to be implemented. In this way, the present invention is not limited to any specific combination of hardware and software.

[0121] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memory, CD-ROM, optical memory, etc.) that contain computer-usable program code.

[0122] The present application is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one or more of the processes or multiple processes and / or blocks Figure 1 one or more of the blocks or multiple blocks.

[0123] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including instruction means, and the instruction means implements the functions specified in Figure 1 one or more of the processes or multiple processes and / or blocks Figure 1 one or more of the blocks or multiple blocks.

[0124] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in Figure 1 one or more of the processes or multiple processes and / or blocks Figure 1 one or more of the blocks or multiple blocks.

[0125] In a typical configuration, a computing device includes one or more processors (CPUs), an input / output interface, a network interface, and a memory.

[0126] The memory may include non-permanent memory in a computer-readable medium, in the form of random access memory (RAM) and / or non-volatile memory such as read-only memory (ROM) or flash memory (flash RAM). The memory is an example of a computer-readable medium.

[0127] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0128] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0129] From the above description, it can be seen that the above embodiments of the present application achieve the following technical effects:

[0130] 1) In the financial service push method of this application, user data is obtained, and user characteristics are extracted based on the user data. Among them, the user data includes user identity information and user consumption information, and the user characteristics represent the consumption characteristics of the user. The user characteristics are analyzed through a user portrait model to obtain a user category prediction result. Among them, the user category prediction result represents the consumption level of the user. The user portrait model is trained through multiple sets of data, and each set of data in the multiple sets of data includes historical user characteristics and the user category prediction results corresponding to the historical user characteristics. A user portrait is generated according to the user category prediction result, and business push is performed on the user according to the user portrait. Among them, the user portrait at least includes the consumption characteristics of the user. Compared with the prior art where financial services cannot be accurately pushed, this application generates a user portrait by extracting user characteristics and performs business push according to the user portrait, so that accurate push can be performed according to the consumption characteristics of the user, improving the user experience. Therefore, the problem of inaccurate push in the prior art can be solved, and the effect of accurately pushing financial services to relevant users can be achieved.

[0131] 2) In the financial service push device of this application, user data is obtained, and user characteristics are extracted based on the user data. Among them, the user data includes user identity information and user consumption information, and the user characteristics represent the consumption characteristics of the user. The user characteristics are analyzed through a user portrait model to obtain a user category prediction result. Among them, the user category prediction result represents the consumption level of the user. The user portrait model is trained through multiple sets of data, and each set of data in the multiple sets of data includes historical user characteristics and the user category prediction results corresponding to the historical user characteristics. A user portrait is generated according to the user category prediction result, and business push is performed on the user according to the user portrait. Among them, the user portrait at least includes the consumption characteristics of the user. Compared with the prior art where financial services cannot be accurately pushed, this application generates a user portrait by extracting user characteristics and performs business push according to the user portrait, so that accurate push can be performed according to the consumption characteristics of the user, improving the user experience. Therefore, the problem of inaccurate push in the prior art can be solved, and the effect of accurately pushing financial services to relevant users can be achieved.

[0132] The above are only the preferred embodiments of this application and are not intended to limit this application. For those skilled in the art, this application can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of this application shall be included within the protection scope of this application.

Claims

1. A method for pushing financial services, characterized in that: include: Acquire user data, and extract user features according to the user data, wherein the user data includes user identity information and user consumption information, and the user features represent the consumption features of the user; Analyze the user features through a user portrait model to obtain a user category prediction result, wherein the user category prediction result represents the consumption level of the user, and the user portrait model is obtained by training multiple sets of data, each of the multiple sets of data includes historical user features and user category prediction results corresponding to the historical user features; A user portrait is generated according to the user category prediction result, and business is pushed to the user according to the user portrait, wherein the user portrait at least includes the consumption characteristics of the user.

2. The method for pushing financial services according to claim 1, characterized in that: Before extracting user features according to the user data, the method further includes: The user data is labeled to obtain a static label and a dynamic label, wherein the static label represents the unchanged user data, and the dynamic label represents the dynamically changing user data; The user data marked with the static tag and the dynamic tag is standardized to obtain standardized user data.

3. The method for pushing financial services according to claim 1, characterized in that: Extracting user features according to the user data includes: Determining the value of the user data through a value assessment model; A one-to-one mapping relationship between the value and the preset user feature is obtained, and the preset user feature corresponding to the value of the user data is determined according to the one-to-one mapping relationship to obtain the user feature.

4. The method for pushing financial services according to claim 1, characterized in that: Pushing services to users based on the user portrait includes: Obtaining the target push product, user geographic location, user consumption frequency, user consumption level, and user interaction times corresponding to the user portrait, wherein the target push product represents a business product whose purchase times by the user are greater than a first preset threshold, and the user interaction times represent the number of times the user has participated in historical push activities; A push mechanism is determined according to the target push product, the user's geographic location, the user's consumption frequency, the user's consumption level, and the user's interaction times, and business products are pushed to the user according to the push mechanism.

5. The method for pushing financial services according to claim 4, characterized in that: Determining a push mechanism according to the target push product, the user's geographical location, the user's consumption frequency, the user's consumption level, and the user's interaction times includes: Pushing the target push product to the user again; Acquire a service product address, calculate an absolute value of a difference between the service product address and the user's geographic location, and push to the user a service product corresponding to the service product address whose absolute value of the difference is less than a second preset threshold; When the user's consumption frequency is greater than the third preset threshold, or the user's consumption level is greater than the fourth preset threshold, or the number of user interactions is greater than the fifth preset threshold, a preferential mechanism is pushed to the user.

6. The method for pushing financial services according to claim 1, characterized in that: Before analyzing the user characteristics through the user portrait model, the method further includes: Constructing an initial user portrait model and obtaining the historical user features, wherein the initial user portrait model includes an input layer, a hidden layer, and an output layer; The historical user features are input into the input layer of the initial user portrait model, and the historical user features are analyzed through the hidden layer until the loss function is minimized to obtain the user portrait model.

7. The method for pushing financial services according to claim 6, characterized in that: Analyzing the historical user characteristics through the hidden layer includes: Through the hidden layer function h (l) =σ(W (l) h (l-1) +b (l) ) analyzes the historical user characteristics, where σ represents the activation function, W (l) represents the weight of the lth layer, b (l) represents the bias of the lth layer, h (l-1) represents the output result of the hidden layer of layer l-1, h (l) Represents the output result of the hidden layer of layer l.

8. A push device for financial services, characterized in that: include: An extraction unit, used to obtain user data, and extract user features according to the user data, wherein the user data includes user identity information and user consumption information, and the user features represent the consumption features of the user; A first analysis unit is configured to analyze the user features through a user portrait model to obtain a user category prediction result, wherein the user category prediction result indicates the consumption level of the user, the user portrait model is obtained by training multiple sets of data, and each set of data in the multiple sets of data includes historical user features and user category prediction results corresponding to the historical user features; A push unit is used to generate a user portrait according to the user category prediction result, and push services to the user according to the user portrait, wherein the user portrait at least includes the consumption characteristics of the user.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the method for pushing financial services as described in any one of claims 1 to 7.

10. An electronic device, characterized in that: include: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the one or more programs include a push method for executing the financial service described in any one of claims 1 to 7.

Images