Bank user-based point exchange mall commodity on-shelf intelligent commodity selection method and system

Abstract

The embodiment of the invention provides an intelligent commodity selection method and system for commodity racking in a point exchange shopping mall based on bank users. According to the embodiment of the invention, the method comprises the steps: constructing a multi-dimensional user preference map through employing the behavior track of a bank user, and generating a dynamic prediction result according to the multi-dimensional user preference map; on the basis of the dynamic prediction result, context-aware calculation is fused, environment variables are calculated, and a commodity recommendation list is generated for each user in a customized mode; utilizing the commodity recommendation list of all bank users to generate an initial commodity selection strategy; implementing shelving according to a commodity shelving scheme in the initial commodity selection strategy, collecting user interaction data after the commodities are shelved, establishing a rapid feedback channel, optimizing the initial commodity selection strategy, and generating an optimal commodity selection strategy; according to the technical scheme provided by the embodiment of the invention, the accuracy of personalized recommendation is improved, and a bank is helped to better cope with market changes, so that the exchange rate and the user satisfaction are improved, the inventory cost is effectively controlled, and the supply chain risk is managed.

Description

Technical Field

[0001] The embodiments of the present application relate to the technical field of financial services, and particularly to an intelligent product selection method and system for commodity listing in an integral redemption mall based on bank users. Background Art

[0002] In modern financial services, bank integral redemption malls have become an important means to attract and retain customers. With the intensification of market competition and the development of technology, banks need more intelligent commodity listing and product selection methods to improve user experience and service quality. Traditional methods based on rules or simple statistical analysis can no longer meet the increasingly complex and personalized user needs;

[0003] Currently, most bank integral redemption malls mainly rely on historical data analysis and simple recommendation algorithms to make commodity listing decisions. These methods usually include user-based collaborative filtering, item-based collaborative filtering, and content recommendation, etc. In addition, some more advanced systems begin to introduce machine learning models to predict users' redemption behaviors and popular commodity categories. However, existing solutions often lack a comprehensive consideration of users' cross-platform behavior trajectories and fail to fully utilize the value of unstructured data such as external market dynamics and social media discussion heat. At the same time, in simulating the market demand response under different listing strategies, existing methods also appear to be insufficiently flexible and accurate. Summary of the Invention

[0004] The embodiments of the present application provide an intelligent product selection method and system for commodity listing in an integral redemption mall based on bank users, so as to solve the problems in the prior art that the recommendation results are not accurate and timely enough, lack an effective feedback mechanism and real-time optimization ability, and are prone to inventory backlog or out-of-stock phenomena.

[0005] In a first aspect, the embodiments of the present application provide an intelligent product selection method for commodity listing in an integral redemption mall based on bank users, including:

[0006] Constructing a multi-dimensional user preference map by using bank users' historical integral redemption behaviors, browsing records, and cross-platform behavior trajectories;

[0007] According to the multi-dimensional user preference map, external market dynamics, and social media discussion heat, combining natural language processing technology to perform sentiment analysis on unstructured data, predicting popular redemption commodity categories and potential change paths of popular redemption commodities, and generating dynamic prediction results;

[0008] Based on the dynamic prediction results, integrating context-aware computing, calculating environmental variables, and customizing a commodity recommendation list for each bank user;

[0009] Using the product recommendation lists of all bank users, the existing inventory status, and the supply chain response time, a reinforcement learning algorithm is adopted to simulate the market demand response under different product listing strategies, determine the optimal product listing plan, and obtain the initial product selection strategy;

[0010] Implement the listing according to the product listing plan in the initial product selection strategy, collect the user interaction data after the product is listed, establish a quick feedback channel, and optimize the initial product selection strategy in real time to generate the optimal product selection strategy.

[0011] Optionally, using the product recommendation lists of all bank users, the existing inventory status, and the supply chain response time, a reinforcement learning algorithm is adopted to simulate the market demand response under different product listing strategies, determine the optimal product listing plan, and obtain the initial product selection strategy, including:

[0012] Using the product recommendation lists of all bank users, the existing inventory status, and the supply chain response time, combined with machine learning techniques, predict the redemption demand for each product within a preset time period to form a comprehensive product information database;

[0013] Based on the data in the comprehensive product information database, combined with the redemption popularity, inventory level, supply chain replenishment cycle, and the life cycle value of bank users, conduct product priority ranking, and select a group of products as the candidate product listing set;

[0014] For the candidate product listing set, use the reinforcement learning algorithm to construct a multi-round simulation environment, introduce virtual agents to represent the redemption behavior patterns of different types of bank users, simulate the market reaction under different product listing strategies, and adjust and optimize the parameters based on the user interaction data after each round of simulation to obtain the optimized simulation results;

[0015] According to the optimized simulation results, evaluate the market demand prediction and potential redemption volume of product combinations under different product listing strategies, screen out the product combinations that meet the preset conditions, and generate a recommended product listing plan;

[0016] Implement the recommended product listing plan in the points redemption mall, collect the user interaction data after the product is listed, and optimize the recommended product listing plan in real time through the quick feedback channel to generate the optimized recommended product listing plan;

[0017] Combining the optimized simulation results and the optimized recommended product listing plan, determine the optimal product listing plan, and establish an intelligent early warning system to monitor the inventory and market dynamics in real time to obtain the initial product selection strategy.

[0018] Optionally, a multi-round simulation environment is constructed using a reinforcement learning algorithm, and virtual agents are introduced to represent the redemption behavior patterns of different types of bank users. The market reaction under different listing strategies is simulated, and parameter adjustment and optimization are performed based on the user interaction data after each round of simulation to obtain optimized simulation results, including:

[0019] Using a reinforcement learning algorithm, according to the product listing cycle within a preset time period, the multi-round simulation environment is constructed and processed. Multiple time slices are divided within each round to simulate the market demand changes at different times, and a multi-round simulation environment is obtained;

[0020] Based on historical redemption behavior, browsing records, cross-platform behavior trajectories, and social network analysis, a preset redemption behavior pattern is constructed, and a group of virtual agents is introduced. The virtual agents are optimized during the simulation process to generate a virtual agent group, and the virtual agent group is deployed into the multi-round simulation environment;

[0021] For each set of candidate listed product sets, multiple different listing strategies are set as input variables, and the multiple different listing strategies are configured in the multi-round simulation environment to generate a configured simulation environment;

[0022] In the configured simulation environment, the virtual agent group interacts according to the preset redemption behavior pattern and the listing strategy of the target round to simulate the actual user redemption behavior, and sentiment analysis is performed on the unstructured user comments through natural language processing technology to generate user interaction data and sentiment tendency scores during the simulation;

[0023] According to the user interaction data and sentiment tendency scores collected after each round of simulation, counterfactual reasoning is applied to evaluate the effects of unimplemented listing strategies, and dynamic adjustment and optimization processing are performed on the parameters of the reinforcement learning model to obtain optimized simulation parameters.

[0024] Optionally, the virtual agent group interacts according to the preset redemption behavior pattern and the listing strategy of the target round to simulate the actual user redemption behavior, and sentiment analysis is performed on the unstructured user comments through natural language processing technology to generate user interaction data and sentiment tendency scores during the simulation, including:

[0025] Using the virtual agent group, deployed into the multi-round simulation environment, the virtual agent group interacts with the listing strategy of the target round. The virtual agent selects products according to the preset redemption behavior pattern and simulates the redemption behavior to generate simulated redemption behavior data;

[0026] Collect the simulated redemption behavior data, record the redemption decisions and redemption results of the virtual agent within each time slice, generate the redemption path and frequency statistics, and obtain the preliminary user interaction data;

[0027] Introduce natural language processing technology to conduct sentiment analysis on the unstructured user comments generated by virtual agents during the simulation process of multi-round simulation environments, identify and quantify the user's sentiment tendency, and evaluate user satisfaction by combining simulation redemption behavior data to generate a sentiment tendency score.

[0028] Optionally, according to the multi-dimensional user preference map, external market dynamics, and social media discussion heat, combine natural language processing technology to conduct sentiment analysis on unstructured data, predict the potential change paths of popular redemption commodity categories and popular redemption commodities, and generate dynamic prediction results, including:

[0029] Utilize the node and edge weight information in the multi-dimensional user preference map to integrate and process the historical point redemption behaviors, browsing records, and cross-platform behavior trajectories of bank users to obtain the user's interest trend and social influence factor;

[0030] Collect and analyze external market dynamic information, identify and process external factors affecting user redemption preferences to obtain an external market influence factor;

[0031] Monitor the discussion heat on social media, use web crawler technology and text mining tools to obtain relevant posts, comments, and shared content to generate a social media discussion dataset;

[0032] Utilize natural language processing technology to conduct sentiment analysis on the social media discussion dataset and other unstructured data, quantify the sentiment tendency of bank users to obtain a sentiment analysis result;

[0033] Combine the user's interest trend, social influence factor, external market influence factor, and sentiment analysis result to construct a comprehensive prediction model, and use the comprehensive prediction model to predict and process the potential change paths of popular redemption commodity categories and popular redemption commodities to generate dynamic prediction results.

[0034] Optionally, implement the product listing according to the product listing plan in the initial product selection strategy, collect the user interaction data after the product is listed, establish a quick feedback channel, and optimize the initial product selection strategy in real time to generate an optimal product selection strategy, including:

[0035] Utilize the product listing plan in the initial product selection strategy to actually arrange and list the products in the point redemption mall to obtain a list of listed products;

[0036] Based on the list of listed products, collect the interaction data between bank users and the listed products through various channels to generate a user interaction dataset;

[0037] Establish a quick feedback channel to transmit the user interaction dataset to the data analysis platform in real time to ensure the immediacy and accuracy of the user interaction dataset;

[0038] Analyze and process the user interaction dataset on the data analysis platform to identify the actual preferences and behavior patterns of bank users, and obtain a user behavior analysis report;

[0039] Based on the user behavior analysis report, combined with context-aware computing, evaluate the effect of the initial product selection strategy, identify areas that need to be adjusted, and generate strategy adjustment suggestions;

[0040] According to the strategy adjustment suggestions, use machine learning algorithms to optimize the initial product selection strategy and obtain the optimized optimal product selection strategy.

[0041] Optionally, use the historical points redemption behavior, browsing records, and cross-platform behavior trajectories of bank users to construct a multi-dimensional user preference map, including:

[0042] Collect the historical points redemption behavior, browsing records, and cross-platform behavior trajectories of users from multiple channels of the banking system to form a user behavior dataset;

[0043] Preprocess and extract features from the user behavior dataset to identify the behavior patterns and preference features of each user, and obtain a user behavior feature library;

[0044] Take each user as a node in the map, and the edges between nodes represent the associations or similarities between users; the weights of the edges are initially set to default values to reflect the basic association degree between users, and an initial user preference map is generated;

[0045] Quantify the interest intensity of users according to the points redemption behavior, browsing frequency, and stay time in the user behavior feature library to obtain user interest scores;

[0046] Combined with social network analysis technology, evaluate the social relationships and influence among users, identify key opinion leaders and tight communities in the social network, and generate social influence factor evaluation results;

[0047] According to the user interest scores and social influence factor evaluation results, dynamically adjust the weights of the edges in the initial user preference map so that the weights of the edges can reflect the interest similarity and social influence between users in real time, and obtain a dynamically updated multi-dimensional user preference map.

[0048] In a second aspect, an embodiment of the present application provides an intelligent product selection system for listing products in a points redemption mall based on bank users, including:

[0049] A construction module for using the historical points redemption behavior, browsing records, and cross-platform behavior trajectories of bank users to construct a multi-dimensional user preference map;

[0050] A prediction module, which is used to perform sentiment analysis on unstructured data according to the multi-dimensional user preference map, external market dynamics, and social media discussion heat, combine natural language processing technology to predict the potential change paths of popular redemption product categories and popular redemption products, and generate dynamic prediction results;

[0051] A calculation module, which is used to calculate environmental variables based on the dynamic prediction results by integrating context-aware computing, and customize a product recommendation list for each bank user;

[0052] A simulation module, which is used to utilize the product recommendation lists of all bank users, the existing inventory status, and the supply chain response time, adopt a reinforcement learning algorithm to simulate the market demand response under different product listing strategies, determine the optimal product listing plan, and obtain an initial product selection strategy;

[0053] An optimization module, which is used to implement product listing according to the product listing plan in the initial product selection strategy, collect user interaction data after product listing, establish a quick feedback channel, and optimize the initial product selection strategy in real time to generate an optimal product selection strategy.

[0054] In a third aspect, an embodiment of the present application provides a computing device, including a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a method for intelligent product selection for product listing in an integral redemption mall based on bank users as described in the first aspect above.

[0055] In a fourth aspect, an embodiment of the present application provides a computer storage medium storing a computer program, and when the computer program is executed by a computer, it implements a method for intelligent product selection for product listing in an integral redemption mall based on bank users as described in the first aspect.

[0056] In the embodiments of the present application, by utilizing the historical points redemption behaviors, browsing records, and cross-platform behavioral trajectories of bank users, a multi-dimensional user preference map is constructed; according to the multi-dimensional user preference map, external market dynamics, and social media discussion heat, combined with natural language processing technology to perform sentiment analysis on unstructured data, predict the potential change paths of popular redemption commodity categories and popular redemption commodities, and generate dynamic prediction results; based on the dynamic prediction results, integrate context-aware computing, calculate environmental variables, and customize a commodity recommendation list for each bank user; utilize the commodity recommendation lists of all bank users, the existing inventory status, and the supply chain response time, adopt a reinforcement learning algorithm to simulate the market demand response under different shelving strategies, determine the optimal commodity shelving plan, and obtain an initial product selection strategy; implement shelving according to the commodity shelving plan in the initial product selection strategy, collect user interaction data after the commodity is shelved, establish a quick feedback channel, and optimize the initial product selection strategy in real time to generate an optimal product selection strategy; the technical solution provided by the present invention improves the accuracy of personalized recommendation, enables the system to predict in advance the popular redemption commodity categories and their potential change paths, helps banks better respond to market changes, thereby improving the redemption rate and user satisfaction, and effectively controlling inventory costs and managing supply chain risks, and enhancing operational efficiency;

[0057] Furthermore, by combining machine learning technology to predict the redemption demand volume of each commodity within a preset time period, a comprehensive commodity information database containing rich information is formed, providing a solid data basis for subsequent product selection decisions; based on the data in the comprehensive commodity information database, considering multiple factors (such as redemption heat, inventory level, supply chain replenishment cycle, and user lifecycle value) to rank the commodities by priority, and selecting a group of commodities as the candidate shelving commodity set, ensuring the scientificity and rationality of product selection; introducing virtual agents to represent the redemption behavior patterns of different types of bank users, simulating the market reaction under different shelving strategies in a multi-round simulation environment, and adjusting and optimizing the parameters according to user interaction data, improving the accuracy and reliability of the simulation results; after implementing the recommended shelving plan to the points redemption mall, collecting user interaction data through the quick feedback channel, optimizing the initial product selection strategy in real time, and establishing an intelligent early warning system to monitor inventory and market dynamics, ensuring the scientificity and forward-looking of the product selection strategy, and reducing the risk of inventory backlog or out-of-stock;

[0058] Furthermore, by interacting the virtual agent group with the listing strategy of the target round, simulating real user redemption behaviors, detailed simulated redemption behavior data is generated, providing an important reference basis for evaluating the effects of different listing strategies; recording the redemption decisions and results of virtual agents in each time slice, forming redemption path and frequency statistics, laying a foundation for subsequent sentiment analysis and user satisfaction evaluation; introducing natural language processing technology to conduct sentiment analysis on the unstructured user comments generated by virtual agents during the simulation process, identifying and quantifying the sentiment tendencies of users, evaluating user satisfaction in combination with the simulated redemption behavior data, generating sentiment tendency scores, and further improving the understanding and prediction capabilities of user behaviors; through meticulous simulation settings and the adaptive learning ability of virtual agents, the authenticity and reliability of the simulation results are improved, ensuring the effectiveness of the final product selection strategy, and at the same time providing more intelligent and personalized service support for the bank.

[0059] These aspects or other aspects of the present application will be more clearly understood in the following description of the embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0061] Figure 1 The flowchart of an intelligent product selection method for listing items in an integral redemption mall based on bank users provided by the present application is shown;

[0062] Figure 2 The structural schematic diagram of an intelligent product selection system for listing items in an integral redemption mall based on bank users provided by the present application is shown;

[0063] Figure 3 The structural schematic diagram of a computing device provided by the present application is shown. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0064] In order to enable those skilled in the art to better understand the solution of the present application, the following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application.

[0065] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit the "first" and "second" to be different types.

[0066] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0067] Figure 1 A flowchart of a method for intelligent product selection based on bank users' points redemption for shopping mall products is provided for the embodiment of the present application, such as Figure 1 As shown, the method includes:

[0068] Step 101: Utilize the historical points redemption behavior, browsing history, and cross-platform behavior tracks of bank users to construct a multi-dimensional user preference map;

[0069] In this step, the multi-dimensional user preference map refers to a complex network structure, in which nodes represent bank users and edge weights are dynamically adjusted to reflect real-time updated user interest strength and social influence factors. Here, "historical points redemption behavior" includes data such as user redemption frequency and redeemed commodity categories, which are used to analyze users' long-term preferences; "browsing history" covers information such as pages visited by users, dwell time, and click paths, which helps understand users' immediate interests; "cross-platform behavior trajectories" refer to user activity records on different platforms (such as bank websites, mobile applications, and offline outlets), which are used to capture users' all-round behavior patterns;

[0070] By collecting and integrating the above three types of data, the system builds a detailed multi-dimensional user preference map. This map can not only show the independent behavioral characteristics of each user, but also reveal the correlation and similarity between users, providing a solid foundation for subsequent personalized recommendations and market trend predictions;

[0071] For example, assume a bank has a large user base. By integrating transaction data from its online mall, mobile banking app, and offline branches, the bank can construct a multi-dimensional user preference map. In this map, user A who often redeems travel-related products and user B who is interested in electronic products form different nodes. Over time, the system dynamically adjusts the weights of the edges based on new redemption behaviors and browsing records, thus reflecting the changing demand of user A who may turn to high-end electronic products.

[0072] Step 102: Based on the multi-dimensional user preference map, external market dynamics, and social media discussion heat, combined with natural language processing technology, conduct sentiment analysis on unstructured data, predict the popular redemption product categories and the potential change paths of popular redemption products, and generate dynamic prediction results;

[0073] In this step, the dynamic prediction results are obtained through comprehensive analysis of the multi-dimensional user preference map, external market dynamics (such as macroeconomic indicators, industry reports), and social media discussion heat. Combining natural language processing technology to conduct sentiment analysis on unstructured data (such as user comments, forum posts) to predict popular redemption product categories and their potential change paths helps to identify market trends in advance and make responses;

[0074] The system uses natural language processing technology to analyze the discussion content on social media, quantify the sentiment tendency of users, and combine the data in the multi-dimensional user preference map to predict which product categories will become popular in the future and the possible development directions of these hotspots;

[0075] Continuing with the above example, when the system detects a large number of positive discussions about environmental protection products on social media, it may predict that the redemption demand for such products will increase significantly in the next few months. Based on this prediction, the bank can prepare more environmental protection-related products in advance in the points redemption mall and push customized recommendations to user groups (such as user A) who show interest in such products.

[0076] Step 103: Based on the dynamic prediction results, integrate context-aware computing, calculate environmental variables, and customize a product recommendation list for each bank user;

[0077] In this step, context-aware computing is a technology that can consider environmental variables (such as time, location, weather, etc.), enabling the system to provide more personalized services according to the current specific situation. Environmental variables refer to factors that affect user choices. For example, seasonal changes may make certain products more popular;

[0078] Based on the dynamic prediction results, the system integrates context-aware computing to calculate environmental variables and customize a product recommendation list for each bank user. This process ensures that the recommended products not only meet the users' immediate needs but also respond to their potential future interests;

[0079] For user A, if the prediction results indicate that environmentally friendly products are about to become popular, and considering that it is currently summer and user A has the habit of outdoor activities, the system will recommend some environmentally friendly products suitable for summer use, such as reusable water bottles or solar chargers. Such recommendations not only meet the users' immediate needs but also guide them to try products of new trends.

[0080] Step 104: Utilize the product recommendation lists of all bank users, the existing inventory status, and the supply chain response time, and adopt a reinforcement learning algorithm to simulate the market demand response under different product listing strategies, determine the optimal product listing plan, and obtain the initial product selection strategy;

[0081] In this step, the reinforcement learning algorithm is a machine learning method that allows the system to optimize the decision-making process through continuous trial and error. The existing inventory status refers to the quantity and location of various products in the warehouse currently; the supply chain response time refers to the time from placing an order to the arrival of the product. Considering these factors comprehensively, the system simulates the market demand response under different product listing strategies to find the optimal solution;

[0082] The system utilizes the product recommendation lists of all bank users, the existing inventory status, and the supply chain response time, adopts a reinforcement learning algorithm to simulate the market demand response under different product listing strategies, and finally determines the optimal product listing plan, that is, the initial product selection strategy;

[0083] Returning to the previous example, the system discovers through simulation that if the inventory of such products is increased and they are placed in a prominent position on the home page before the upcoming environmentally friendly product boom, the redemption rate can be maximized. Therefore, the bank decides to adjust its product layout to ensure that environmentally friendly products are fully displayed and the inventory is replenished in a timely manner to cope with the expected demand growth.

[0084] Step 105: Implement the product listing according to the product listing plan in the initial product selection strategy, collect the user interaction data after the product listing, establish a rapid feedback channel, and optimize the initial product selection strategy in real time to generate the optimal product selection strategy;

[0085] In this step, the rapid feedback channel is a mechanism that allows the system to quickly collect and analyze the user interaction data (such as click-through rate, redemption rate, user reviews) after the product listing, so as to adjust the product selection strategy in real time. Through this continuous optimization process, the system can maintain the effectiveness and forward-looking nature of the product selection strategy;

[0086] After the products are listed according to the product listing plan in the initial product selection strategy, the system establishes a quick feedback channel to collect user interaction data after the products are listed, optimize the initial product selection strategy in real time, and finally generate the optimal product selection strategy.

[0087] For example, after an environmental protection product is listed, the system monitors that user A shows extremely high interest in a specific environmental protection water bottle, but the redemption volume does not reach the expected level. After further analysis, it is found that the reason may be that the price is slightly high. So, the bank quickly adjusted the promotion strategy and reduced the price of this water bottle, and as a result, the redemption volume increased significantly. In this way, the system continuously optimizes the product selection strategy to ensure that each adjustment can bring a better user experience and a higher redemption rate.

[0088] By implementing the above steps, the present invention provides a highly intelligent and flexible product listing and selection method, which can comprehensively capture changes in user preferences, accurately predict market trends, and dynamically optimize product recommendation and listing strategies. Specifically, constructing a multi-dimensional user preference map improves the accuracy of personalized recommendations; generating dynamic prediction results enables the bank to respond to market changes in advance; customizing the generation of product recommendation lists enhances the user experience; determining the optimal product listing plan ensures operational efficiency; and real-time optimizing the product selection strategy ensures that the product selection strategy is always in the best state. The entire process is interlocked, not only improving the operational efficiency and service quality of the bank's points redemption mall, but also bringing a more personalized and satisfactory redemption experience to users.

[0089] In order to solve the problem that the product selection strategy in the prior art is not flexible and accurate enough and improve the effect of market demand prediction and inventory management in the prior art, in one embodiment, according to step 104, using the product recommendation lists of all bank users, the existing inventory status, and the supply chain response time, the reinforcement learning algorithm is used to simulate the market demand response under different listing strategies, determine the optimal product listing plan, and obtain the initial product selection strategy, specifically including:

[0090] Using the product recommendation lists of all bank users, the existing inventory status, and the supply chain response time, combined with machine learning techniques, predict the redemption demand for each product within a preset time period to form a comprehensive product information database; based on the data in the comprehensive product information database, combined with the redemption popularity, inventory level, supply chain replenishment cycle, and the life cycle value of bank users, conduct product priority ranking, and select a group of products as the candidate product set for listing; for the candidate product set for listing, use the reinforcement learning algorithm to construct a multi-round simulation environment, and introduce virtual agents to represent the redemption behavior patterns of different types of bank users, simulate the market reaction under different listing strategies, and adjust and optimize the parameters based on the user interaction data after each round of simulation to obtain the optimized simulation results; according to the optimized simulation results, evaluate the market demand prediction and potential redemption volume of the product portfolio under different listing strategies, screen out the product portfolio that meets the preset conditions, and generate a recommended listing plan; implement the recommended listing plan to the points redemption mall, and collect the user interaction data after the product is listed, and optimize the recommended listing plan in real time through the fast feedback channel to generate the optimized recommended listing plan;

[0091] Combined with the optimized simulation results and the optimized recommended listing plan, determine the optimal product listing plan, and establish an intelligent early warning system to monitor the inventory and market dynamics in real time to obtain the initial product selection strategy;

[0092] In this embodiment, the comprehensive product information database includes data such as the product recommendation lists of all bank users, the existing inventory status, and the supply chain response time. These data are used to predict the redemption demand for each product within a preset time period in combination with machine learning techniques. In this way, the system can form a comprehensive product information database with rich information, providing a solid data foundation for subsequent product priority ranking and product selection decisions;

[0093] In the embodiments of the present application, first, the system uses the product recommendation lists of all bank users, the existing inventory status, and the supply chain response time, and combines machine learning techniques to predict the redemption demand for each product within a preset time period, forming a comprehensive product information database. Next, based on the data in the comprehensive product information database, the system considers the redemption popularity of the product, the inventory level, the supply chain replenishment cycle, and the customer lifetime value (CLV) of bank users to rank the products by priority, and selects a group of products as the candidate product set for listing. Subsequently, for the candidate product set for listing, the system uses a reinforcement learning algorithm to construct a multi-round simulation environment, and introduces virtual agents to represent the redemption behavior patterns of different types of bank users to simulate the market reaction under different listing strategies. After each round of simulation ends, the system adjusts and optimizes the parameters based on the user interaction data to obtain the optimized simulation results. According to the optimized simulation results, the system evaluates the market demand prediction and potential redemption volume of the product portfolio under different listing strategies, filters out the product portfolios that meet the preset conditions, and generates a recommended listing plan. Finally, the recommended listing plan is implemented in the points redemption mall, and the user interaction data after the product is listed is collected, and the recommended listing plan is optimized in real time through a fast feedback channel to generate an optimized recommended listing plan; combining the optimized simulation results and the optimized recommended listing plan, the optimal product listing plan is determined, and an intelligent early warning system is established to monitor the inventory and market dynamics in real time, and finally the initial product selection strategy is obtained;

[0094] For example, in a practical application, assume that a certain bank plans to launch a series of new environmental protection products to meet the upcoming market trends. By collecting and analyzing the product recommendation lists of all bank users, the existing inventory status, and the supply chain response time, the system can predict the redemption demand for various products in the coming weeks. For those products with high expected demand, such as environmental protection water bottles or solar chargers, the system will prioritize them in the candidate product set for listing. Then, by constructing a multi-round simulation environment, the system simulates the impact of different listing strategies (such as promotion activity arrangements, product display positions, etc.) on the market reaction. During this process, the virtual agents imitate the redemption behavior patterns of different types of bank users (such as young white-collar workers, housewives, etc.), making the simulation more realistic. After multiple rounds of simulation and optimization, the system finally determines an optimal product listing plan, which not only ensures the sufficient supply of high-demand products but also effectively controls the inventory cost. In addition, by monitoring the user interaction data and market dynamics in real time, the bank can quickly adjust its strategy to cope with any emergencies and ensure that the product selection strategy is always in the best state.

[0095] To address the problem of inaccurate and inflexible product selection strategy simulation in the prior art and improve the effect of market demand prediction and optimization in the prior art, as described in the previous embodiment, for the candidate product set to be listed, a multi-round simulation environment is constructed using a reinforcement learning algorithm, and virtual agents are introduced to represent the redemption behavior patterns of different types of bank users, simulate the market reaction under different listing strategies, and perform parameter adjustment and optimization based on the user interaction data after each round of simulation to obtain an optimized simulation result, specifically including:

[0096] Using a reinforcement learning algorithm, according to the product listing cycle within a preset time period, construct and process a multi-round simulation environment. Divide multiple time slices within each round to simulate the market demand changes at different time periods, and obtain a multi-round simulation environment; Based on historical redemption behavior, browsing records, cross-platform behavior trajectories, and social network analysis, construct a preset redemption behavior pattern, introduce a group of virtual agents, use the virtual agents to optimize during the simulation process, generate a virtual agent group, and deploy the virtual agent group into the multi-round simulation environment; For each set of candidate product sets to be listed, set multiple different listing strategies as input variables, configure the multiple different listing strategies in the multi-round simulation environment, and generate a configured simulation environment; In the configured simulation environment, enable the virtual agent group to interact with the listing strategy of the target round according to the preset redemption behavior pattern, simulate the actual user redemption behavior, and perform sentiment analysis on unstructured user comments through natural language processing technology to generate user interaction data and sentiment tendency scores during the simulation; According to the user interaction data and sentiment tendency scores collected after each round of simulation, apply counterfactual reasoning to evaluate the effect of unimplemented listing strategies, and perform dynamic adjustment and optimization processing on the parameters of the reinforcement learning model to obtain optimized simulation parameters;

[0097] In this embodiment, the multi-round simulation environment refers to multiple simulation scenarios constructed by a reinforcement learning algorithm according to the product listing cycle within a preset time period. Multiple time slices are divided within each round to finely simulate the market demand changes at different time periods. This setting enables the system to capture demand fluctuations at different times of the day (such as morning rush hour, evening rush hour) or different days of the week (such as weekdays and weekends). The virtual agent group is a group of agents with adaptive learning ability constructed by analyzing historical redemption behavior, browsing records, cross-platform behavior trajectories, and social network discussion heat. Each agent represents a type of bank user with a specific redemption behavior pattern. These agents can self-evolve according to feedback during the simulation process to generate more realistic user interaction behaviors;

[0098] In the embodiments of the present application, first, the system uses a reinforcement learning algorithm to construct a multi-round simulation environment based on the product listing cycle within a preset time period. Multiple time slices are divided within each round to simulate the market demand changes at different time periods, so as to obtain a multi-round simulation environment. Then, based on historical redemption behaviors, browsing records, cross-platform behavior trajectories, and social network analysis, the system constructs a preset redemption behavior pattern, introduces a group of virtual agents to form a virtual agent group, and deploys the virtual agent group into the multi-round simulation environment. For each set of candidate listed product sets, the system sets multiple different listing strategies as input variables, configures these strategies in the multi-round simulation environment, and generates a configured simulation environment. Then, in the configured simulation environment, the virtual agent group interacts according to the preset redemption behavior pattern and the listing strategy of the target round, simulates the actual user redemption behavior, and performs sentiment analysis on unstructured user comments through natural language processing technology to generate user interaction data and sentiment tendency scores during the simulation. Finally, according to the user interaction data and sentiment tendency scores collected after each round of simulation, counterfactual reasoning is applied to evaluate the effects of unimplemented listing strategies, and the parameters of the reinforcement learning model are dynamically adjusted and optimized to obtain optimized simulation parameters;

[0099] For example, in a practical application, assume that a certain bank plans to launch a series of new environmental protection products to meet the upcoming market trends. By constructing a multi-round simulation environment, the system can simulate the market demand changes at different time periods (such as weekdays and weekends, day and night). Based on the detailed behavior data analysis of bank users, the system generates a group of virtual agent groups, and each agent represents a user group with a specific redemption behavior pattern. For sets of candidate listed products such as newly launched environmental protection water bottles and solar chargers, the system sets multiple different listing strategies (such as promotion activity arrangements, product display positions, etc.) and configures them in the multi-round simulation environment. During the simulation, the virtual agents interact according to the preset redemption behavior pattern and the listing strategy, simulate the real user redemption behavior, and perform sentiment analysis on the user comments generated by the simulation through natural language processing technology to generate detailed user interaction data and sentiment tendency scores. After each round of simulation ends, the system applies counterfactual reasoning to evaluate the effects of unimplemented strategies, dynamically adjusts and optimizes the parameters of the reinforcement learning model to ensure that the final simulation results are as close as possible to the actual situation. This process not only improves the accuracy of the simulation, but also enables the bank to quickly adjust its strategies to cope with any emergencies and ensure that the product selection strategy is always in the best state.

[0100] To address the issues of insufficient authenticity in simulating user behavior and inadequate sentiment analysis in the prior art, as well as to improve the effectiveness of product selection strategy optimization in the prior art, as described in the previous embodiment, in the configured simulation environment, the virtual agent group interacts with the listing strategy of the target round according to the preset redemption behavior pattern, simulating actual user redemption behavior, and performing sentiment analysis on unstructured user comments through natural language processing technology to generate user interaction data and sentiment tendency scores during the simulation, specifically including:

[0101] Deploy the virtual agent group into a multi-round simulation environment, enabling the virtual agent group to interact with the listing strategy of the target round. The virtual agent selects products according to the preset redemption behavior pattern and simulates the redemption behavior to generate simulated redemption behavior data; collect the simulated redemption behavior data, record the redemption decisions and results of the virtual agent in each time slice, generate redemption paths and frequency statistics, and obtain preliminary user interaction data; introduce natural language processing technology to perform sentiment analysis on the unstructured user comments generated by the virtual agent during the simulation process of the multi-round simulation environment, identify and quantify the user's sentiment tendency, and evaluate user satisfaction in combination with the simulated redemption behavior data to generate sentiment tendency scores;

[0102] In this embodiment, the simulated redemption behavior data refers to the data generated when the virtual agent group selects products and simulates the redemption behavior according to the preset redemption behavior pattern in the multi-round simulation environment. These data record the redemption decisions and results of the virtual agent in each time slice, forming detailed redemption paths and frequency statistics, which are the basis of the preliminary user interaction data. The sentiment tendency score is the result obtained by introducing natural language processing technology to perform sentiment analysis on the unstructured user comments generated by the virtual agent during the simulation process. It not only identifies the user's sentiment tendency (such as positive, negative, or neutral), but also evaluates user satisfaction in combination with the simulated redemption behavior data, providing a more comprehensive user experience feedback;

[0103] In the embodiments of the present application, first, the system deploys a virtual agent group into a multi-round simulation environment, enabling the virtual agent group to interact with the listing strategy of the target round. The virtual agent selects products according to a preset redemption behavior pattern, simulates the redemption behavior, and generates simulated redemption behavior data. Next, the system collects the simulated redemption behavior data, details the redemption decisions and results of the virtual agent in each time slice, forms a redemption path and frequency statistics, and obtains preliminary user interaction data. Subsequently, the system introduces natural language processing technology to perform sentiment analysis on the unstructured user comments generated by the virtual agent during the simulation process, and identifies and quantifies the user's sentiment tendency. Finally, the user satisfaction is evaluated in combination with the simulated redemption behavior data to generate a sentiment tendency score. This process ensures that the simulation results not only reflect the user's redemption behavior but also capture their emotional responses, thereby providing richer information support for subsequent strategy optimization;

[0104] For example, in a practical application, assume that a certain bank plans to launch a series of new environmental protection products to meet the upcoming market trends. In the configured simulation environment, the virtual agent group represents different types of bank users (such as young white-collar workers, housewives, etc.), and they interact with different listing strategies according to a preset redemption behavior pattern to simulate real redemption behaviors. During this process, the virtual agent selects interesting environmental protection water bottles or solar chargers, simulates the entire process from browsing to redemption, and generates detailed simulated redemption behavior data. At the same time, the virtual agent also simulates and generates some unstructured user comments, such as "The design of this environmental protection water bottle is very fashionable!" or "The price of the solar charger is a bit high". The system performs sentiment analysis on these comments through natural language processing technology, identifies positive and negative sentiment tendencies, and evaluates the user satisfaction in combination with the simulated redemption behavior data, and finally generates a sentiment tendency score. This detailed simulation and analysis help the bank better understand the preferences and dissatisfaction of potential users, enabling the bank to adjust the product selection strategy targeted to ensure that the launched environmental protection products not only meet the market demand but also gain high recognition from users.

[0105] To solve the problems of inaccurate market trend prediction and untimely capture of changes in user preferences in the prior art and improve the effect of product selection strategy optimization in the prior art, as another embodiment, according to what is described in step 102, based on the multi-dimensional user preference map and external market dynamics and social media discussion heat, combined with natural language processing technology to perform sentiment analysis on unstructured data, predict the potential change paths of popular redemption product categories and popular redemption products, and generate dynamic prediction results, specifically including:

[0106] Integrate and process the historical points redemption behaviors, browsing records, and cross-platform behavioral trajectories of bank users by using the node and edge weight information in the multi-dimensional user preference graph to obtain the user's interest trend and social influence factor; collect and analyze external market dynamic information, identify and process the external factors affecting the user's redemption preference to obtain the external market influence factor; monitor the discussion heat on social media, use web crawler technology and text mining tools to obtain relevant posts, comments, and shared content, and generate a social media discussion dataset; use natural language processing technology to perform sentiment analysis on the social media discussion dataset and other unstructured data, quantify the sentiment tendency of bank users to obtain the sentiment analysis result; combine the user's interest trend, social influence factor, external market influence factor, and sentiment analysis result to construct a comprehensive prediction model, and use the comprehensive prediction model to predict and process the potential change paths of popular redemption commodity categories and popular redemption commodities to generate a dynamic prediction result;

[0107] In this embodiment, the interest trend refers to the long-term preferences and development directions of users obtained by integrating the historical points redemption behaviors, browsing records, and cross-platform behavioral trajectories of bank users. This information is used to identify the immediate and potential needs of users. The social influence factor is the association strength and social influence among users reflected by the node and edge weight information in the multi-dimensional user preference graph, which helps the system understand the interaction among user groups and its impact on individual preferences. The external market influence factor is the result obtained by identifying and processing external factors (such as macroeconomic indicators, industry reports, competitor activities, etc.) that affect the user's redemption preference, and these factors may indirectly affect the user's redemption decision. The social media discussion dataset is a collection of relevant posts, comments, and shared content obtained from social media by using web crawler technology and text mining tools, which reflects the discussion heat and viewpoints of the public on specific commodities or categories. The sentiment analysis result is the output after quantifying the sentiment tendency of the social media discussion dataset and other unstructured data by using natural language processing technology, which is used to evaluate the impact of market sentiment and social public opinion;

[0108] In the embodiments of the present application, first, the system uses the node and edge weight information in the multi-dimensional user preference map to integrate and process the historical points redemption behaviors, browsing records, and cross-platform behavior trajectories of bank users, obtaining the user's interest trends and social influence factors. Next, the system collects and analyzes external market dynamic information, identifies and processes the external factors affecting the user's redemption preferences, and obtains the external market influence factors. Subsequently, the system monitors the discussion heat on social media, uses web crawler technology and text mining tools to obtain relevant posts, comments, and shared content, and generates a social media discussion dataset. Then, the system uses natural language processing technology to perform sentiment analysis on the social media discussion dataset and other unstructured data, quantifies the sentiment tendency of bank users, and obtains the sentiment analysis results. Finally, the system combines the user's interest trends, social influence factors, external market influence factors, and sentiment analysis results to construct a comprehensive prediction model, and uses this model to predict and process the potential change paths of popular redemption product categories and popular redemption products, generating dynamic prediction results. This process ensures that the prediction is not only based on the user's historical behavior and current preferences, but also fully considers the changes in the external market environment and social opinion, thereby improving the accuracy and foresight of the prediction;

[0109] For example, in a practical application, assume that a certain bank hopes to predict in advance the market demand changes of environmental protection products. First, by integrating the user's historical points redemption behaviors, browsing records, and cross-platform behavior trajectories, the system determines that certain user groups show a strong interest trend in environmental protection products, and these users form a close social circle in the network and have significant social influence on each other. Next, the system analyzes the external market dynamic information and finds that there have been frequent news about environmental protection policies recently, which may be an important external factor affecting the user's redemption preferences. At the same time, by monitoring the discussion heat on social media, the system notices that the positive evaluations and recommendations for environmental protection products are increasing continuously, especially the interest in portable solar chargers has increased significantly. After using natural language processing technology to perform sentiment analysis on these discussion contents, the system quantifies the sentiment tendency of users and confirms the high attention of the market to environmental protection products. Finally, the system combines all this information to construct a comprehensive prediction model, predicting that the demand for environmental protection products will increase significantly in the next few months, especially portable solar chargers will become the new hot spot. Based on this, the bank adjusts the product layout of its points redemption mall, prepares sufficient inventory of environmental protection products in advance, and launches customized recommendations for user groups showing interest in such products, successfully improving the redemption rate and user satisfaction.

[0110] To address the problem of the lack of real-time feedback and optimization mechanisms after the implementation of the product selection strategy in the prior art, and to improve the user satisfaction and redemption rate in the prior art, as another embodiment, as described in step 105, implement the product listing according to the product listing plan in the initial product selection strategy, collect the user interaction data after the product listing, establish a quick feedback channel, and perform real-time optimization on the initial product selection strategy to generate the optimal product selection strategy, specifically including:

[0111] Utilize the product listing plan in the initial product selection strategy to actually arrange and list the products in the points redemption mall to obtain a list of listed products; based on the list of listed products, collect the interaction data of bank users with the listed products through various channels to generate a user interaction data set; establish a quick feedback channel to transmit the user interaction data set to the data analysis platform in real time, so that the user interaction data set has instantaneity and accuracy; use the data analysis platform to analyze and process the user interaction data set, identify the actual preferences and behavior patterns of bank users, and obtain a user behavior analysis report; based on the user behavior analysis report, combine context-aware computing, evaluate the effect of the initial product selection strategy, identify areas that need to be adjusted, and generate strategy adjustment suggestions; according to the strategy adjustment suggestions, adopt machine learning algorithms to optimize the initial product selection strategy to obtain the optimized optimal product selection strategy;

[0112] In this embodiment, the list of listed products refers to the product list obtained after actually arranging and listing the products in the points redemption mall according to the product listing plan in the initial product selection strategy. These products are carefully selected and arranged to maximize the attraction of users' interest and participation. The user interaction data set is a collection formed by collecting the interaction data of bank users with the listed products through various channels (such as website click records, APP operation logs, customer service feedback, etc.) to evaluate the actual preferences and behavior patterns of users. The quick feedback channel is a real-time transmission mechanism that ensures that the user interaction data set can be quickly and accurately transmitted to the data analysis platform for instant analysis and response. The user behavior analysis report is the result generated after in-depth analysis of the user interaction data set, which reveals the actual preferences, behavior patterns of users and their interest levels in different products. The strategy adjustment suggestions are specific improvement measures proposed based on the user behavior analysis report and combined with context-aware computing to evaluate the effect of the initial product selection strategy to optimize the product selection strategy. The machine learning algorithm specifically refers to using advanced algorithms to iteratively optimize the initial product selection strategy, continuously update the parameters of the product selection model to reflect the latest market dynamics and changes in user preferences;

[0113] In the embodiments of the present application, first, the system uses the product listing plan in the initial product selection strategy to actually arrange and list the products in the points redemption mall, obtaining a list of listed products. Next, based on the list of listed products, the system collects interaction data between bank users and the listed products through multiple channels, forming a user interaction data set. To ensure the timeliness and accuracy of the data, the system establishes a fast feedback channel to transmit the user interaction data set to the data analysis platform in real time. Subsequently, on the data analysis platform, the system analyzes and processes the user interaction data set, identifies the actual preferences and behavior patterns of bank users, and generates a user behavior analysis report. Based on this report, the system combines context-aware computing, evaluates the effect of the initial product selection strategy, identifies areas that need to be adjusted, and generates strategy adjustment suggestions. Finally, according to the strategy adjustment suggestions, the system uses machine learning algorithms to optimize the initial product selection strategy, continuously updates the product selection model parameters, and finally obtains an optimized optimal product selection strategy;

[0114] For example, in a practical application, assume that a certain bank launches a batch of environmental protection products in the points redemption mall according to the initial product selection strategy, including portable solar chargers and reusable water bottles. After the products are listed, the system collects interaction data between users and these environmental protection products through multiple channels (such as website click records, APP operation logs, customer service feedback, etc.), forming a detailed user interaction data set. To ensure the timeliness and accuracy of the data, the system establishes a fast feedback channel so that every user interaction can be captured in a timely manner and transmitted to the data analysis platform. The data analysis platform conducts in-depth analysis on these interaction data and finds that although most users show strong interest in environmental protection products, some users have concerns about the price of solar chargers. Based on this analysis result, the system generates a user behavior analysis report, indicating that price sensitivity may be one of the key factors affecting the redemption decision. According to this report, the system proposes strategy adjustment suggestions, such as time-limited discounts or combined package offers, to lower the psychological threshold of users. By using machine learning algorithms to optimize the initial product selection strategy, the bank not only adjusts the pricing strategy of environmental protection products but also launches promotional activities in a timely manner, significantly improving the redemption rate and user satisfaction. This real-time feedback and optimization mechanism ensures that the product selection strategy is always in the best state, effectively improving the operation efficiency and service quality of the bank's points redemption mall.

[0115] To solve the problems of insufficient comprehensiveness in user preference modeling and untimely real-time updates in the prior art and to improve the effect of personalized recommendation and service quality in the prior art, as another embodiment, according to what is described in step 101, a multi-dimensional user preference map is constructed using the historical points redemption behaviors, browsing records, and cross-platform behavior trajectories of bank users, specifically including:

[0116] Collect the historical points redemption behaviors, browsing records, and cross-platform behavior trajectories of users from multiple channels of the banking system to form a user behavior dataset; preprocess and extract features from the user behavior dataset to identify the behavior patterns and preference features of each user, and obtain a user behavior feature library; regard each user as a node in the graph, and the edges between the nodes represent the association or similarity between users; the weight of the edge is initially set to the default value, which is used to reflect the basic association degree between users, and generate an initial user preference graph; quantify the interest intensity of users according to the points redemption behavior, browsing frequency, and stay time in the user behavior feature library to obtain a user interest score; combine social network analysis techniques to evaluate the social relationships and influence among users, identify the key opinion leaders and tight communities in the social network, and generate the evaluation results of social influence factors; according to the user interest score and the evaluation results of social influence factors, dynamically adjust the weights of the edges in the initial user preference graph so that the weights of the edges can reflect the interest similarity and social influence between users in real time, and obtain a dynamically updated multi-dimensional user preference graph;

[0117] In this embodiment, the user behavior dataset refers to the historical points redemption behaviors, browsing records, and cross-platform behavior trajectories of users collected from multiple channels of the banking system (such as websites, mobile applications, offline branches, etc.). These data are used to comprehensively capture the interaction patterns and preference features of users. The user behavior feature library is the result obtained by preprocessing and extracting features from the user behavior dataset. It identifies the behavior patterns and preference features of each user and provides a basis for the subsequent construction of the preference graph. The initial user preference graph regards each user as a node in the graph, and the edges between the nodes represent the association or similarity between users. The weight of the edge is initially set to the default value, which reflects the basic association degree between users. The user interest score measures the interest intensity of users by quantifying the points redemption behavior, browsing frequency, and stay time in the user behavior feature library, providing a quantitative evaluation of the current preferences of users. The evaluation results of social influence factors are the output after combining social network analysis techniques to evaluate the social relationships and influence among users, helping to identify the key opinion leaders (KOLs) and tight communities in the social network. The dynamically updated multi-dimensional user preference graph refers to dynamically adjusting the weights of the edges according to the user interest score and the evaluation results of social influence factors, so that the graph can reflect the interest similarity and social influence between users in real time;

[0118] In the embodiments of the present application, first, the system collects the historical points redemption behaviors, browsing records, and cross-platform behavior trajectories of users from multiple channels of the banking system to form a user behavior dataset. Next, the system preprocesses and extracts features from the user behavior dataset to identify the behavior patterns and preference features of each user, obtaining a user behavior feature library. Then, the system takes each user as a node in the graph, and the edges between the nodes represent the associations or similarities between users. The weights of the edges are initially set to default values to generate an initial user preference graph. Subsequently, the system quantifies the interest intensity of users based on the points redemption behaviors, browsing frequencies, and residence times in the user behavior feature library to obtain user interest scores. At the same time, the system combines social network analysis techniques to evaluate the social relationships and their influences between users, identifies the key opinion leaders and tight communities in the social network, and generates the evaluation results of social influence factors. Finally, based on the user interest scores and the evaluation results of social influence factors, the system dynamically adjusts the weights of the edges in the initial user preference graph so that the weights of the edges can reflect the interest similarity and social influence between users in real time, and finally obtains a dynamically updated multi-dimensional user preference graph;

[0119] For example, in a practical application, assume that a certain bank hopes to improve the user experience and redemption rate of its points redemption mall. First, the system collects the historical points redemption behaviors, browsing records, and cross-platform behavior trajectories of users from multiple channels of the bank (such as online shopping malls, mobile banking APPs, offline branches, etc.) to form a detailed user behavior dataset. Next, the system preprocesses and extracts features from these data to identify the behavior patterns and preference features of each user, and establishes a user behavior feature library. Based on this feature library, the system constructs an initial user preference graph, in which each user is represented as a node, and the edges between the nodes reflect the basic association degree between users. To more accurately capture the interest changes of users, the system quantifies the interest intensity of users based on factors such as points redemption behaviors, browsing frequencies, and residence times to obtain user interest scores. In addition, the system also combines social network analysis techniques to evaluate the social relationships and their influences between users, and identifies the key opinion leaders and tight communities in the social network. According to these evaluation results, the system dynamically adjusts the weights of the edges in the initial user preference graph to ensure that the graph can reflect the interest similarity and social influence between users in real time. This detailed user preference modeling not only improves the accuracy of personalized recommendations, but also enhances user participation and satisfaction, and ultimately promotes a higher redemption rate and better user experience.

[0120] In summary, the present invention significantly improves the operational efficiency and service quality of the bank's points redemption mall by constructing a multi-dimensional user preference map, dynamically predicting market trends, customizing the generation of a product recommendation list, simulating the market demand response under different listing strategies, and real-time optimizing the product selection strategy. Specifically, the system comprehensively captures the interest changes and social influences of users, accurately predicts popular products and their potential change paths, ensuring that personalized recommendations not only meet the immediate needs of users but also respond to future potential interests. By simulating and optimizing the listing strategy through a reinforcement learning algorithm, the system can effectively control inventory costs and manage supply chain risks, while establishing a rapid feedback channel to continuously optimize the product selection strategy and ensure it is always in the best state. Ultimately, the present invention not only increases the redemption rate and satisfaction of users but also enhances the flexibility and adaptability of the bank in a complex and changing market environment, providing more intelligent and personalized service support for the bank.

[0121] Figure 2 FIG. provides a schematic structural diagram of an intelligent product selection device (or system) for product listing in a points redemption mall based on bank users, as Figure 2 shown, the device includes:

[0122] A construction module 21, configured to construct a multi-dimensional user preference map by using the historical points redemption behaviors, browsing records, and cross-platform behavioral trajectories of bank users;

[0123] A prediction module 22, configured to predict the popular redemption product categories and the potential change paths of popular redemption products according to the multi-dimensional user preference map, external market dynamics, and social media discussion heat, and perform sentiment analysis on unstructured data by combining natural language processing technology to generate a dynamic prediction result;

[0124] A calculation module 23, configured to customize and generate a product recommendation list for each bank user by fusing context-aware calculation and calculating environmental variables based on the dynamic prediction result;

[0125] A simulation module 24, configured to use the product recommendation lists of all bank users, the existing inventory status, and the supply chain response time, and adopt a reinforcement learning algorithm to simulate the market demand response under different listing strategies, determine the optimal product listing plan, and obtain an initial product selection strategy;

[0126] An optimization module 25, configured to implement listing according to the product listing plan in the initial product selection strategy, collect user interaction data after product listing, establish a rapid feedback channel, and perform real-time optimization on the initial product selection strategy to generate an optimal product selection strategy.

[0127] Figure 2 The described intelligent product selection device for product listing in a points redemption mall based on bank users can execute Figure 1A method for intelligent product selection for commodity listing in an integral exchange mall based on bank users described in the illustrated embodiment, its implementation principle and technical effects will not be elaborated further. For the intelligent product selection device for commodity listing in the integral exchange mall based on bank users in the above embodiment, the specific ways in which each module and unit perform operations have been described in detail in the embodiment related to this method, and will not be elaborated here.

[0128] In a possible design, Figure 2 An intelligent product selection device for commodity listing in an integral exchange mall based on bank users in the illustrated embodiment can be implemented as a computing device, such as Figure 3 shown, this computing device may include a storage component 31 and a processing component 32;

[0129] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32.

[0130] The processing component 32 is used for the Figure 1 A method for intelligent product selection for commodity listing in an integral exchange mall based on bank users in the above embodiment.

[0131] Among them, the processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented by one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components for executing the above method.

[0132] The storage component 31 is configured to store various types of data to support operations on the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0133] Of course, the computing device may necessarily also include other components, such as input / output interfaces, display components, communication components, etc.

[0134] The input / output interface provides an interface between the processing component and the peripheral interface module, and the above peripheral interface module may be an output device, an input device, etc.

[0135] The communication component is configured to facilitate communication between the computing device and other devices in a wired or wireless manner, etc.

[0136] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. At this time, the computing device can refer to a cloud server. The above processing component, storage component, etc. can be basic server resources rented or purchased from a cloud computing platform.

[0137] The embodiment of the present application also provides a computer storage medium storing a computer program, and when the computer program is executed by a computer, it can implement the above Figure 1 shown embodiment of an intelligent product selection method for listing products in an integral redemption mall based on bank users.

[0138] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the above-described systems, devices, and units can refer to the corresponding processes in the foregoing method embodiments, and will not be elaborated herein.

[0139] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative labor.

[0140] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disc, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0141] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of each embodiment of the present application.

Claims

1. A method for intelligent product selection based on bank users' points redemption for shopping mall products, characterized in that: include: Utilize bank users’ historical points redemption behavior, browsing history, and cross-platform behavior trajectories to build a multi-dimensional user preference map; Based on multi-dimensional user preference maps, external market dynamics, and social media discussion heat, we use natural language processing technology to conduct sentiment analysis on unstructured data, predict the categories of popular redemption products and the potential change paths of popular redemption products, and generate dynamic prediction results; Based on the dynamic prediction results, context-aware computing is integrated to calculate environmental variables, and a product recommendation list is customized for each bank user; Using the product recommendation list, current inventory status, and supply chain response time of all bank users, a reinforcement learning algorithm is used to simulate the market demand response under different listing strategies, determine the optimal product listing plan, and obtain the initial product selection strategy; The product listing plan in the initial product selection strategy is implemented, and user interaction data after the product is listed is collected. A fast feedback channel is established to optimize the initial product selection strategy in real time and generate the optimal product selection strategy.

2. The method according to claim 1, characterized in that Using the product recommendation list, current inventory status, and supply chain response time of all bank users, a reinforcement learning algorithm is used to simulate the market demand response under different listing strategies, determine the optimal product listing plan, and obtain the initial product selection strategy, including: Using the product recommendation list, current inventory status, and supply chain response time of all bank users, combined with machine learning technology, the exchange demand for each product within a preset time period is predicted to form a comprehensive product information database; Based on the data of the comprehensive product information database, combined with the redemption popularity, inventory level, supply chain replenishment cycle and bank user life cycle value of the product, the product priority is sorted, and a group of products are selected as the candidate product set for listing; For the candidate product set, a multi-round simulation environment is constructed using a reinforcement learning algorithm. Virtual agents are introduced to represent the redemption behavior patterns of different types of bank users to simulate the market response under different listing strategies. Parameters are adjusted and optimized based on the user interaction data after each round of simulation to obtain the optimized simulation results. Based on the optimized simulation results, the market demand forecast and potential redemption volume of product combinations under different listing strategies are evaluated, product combinations that meet the preset conditions are screened out, and recommended listing plans are generated; Implement the recommended listing plan to the points redemption mall, collect user interaction data after the product is listed, optimize the recommended listing plan in real time through the fast feedback channel, and generate an optimized recommended listing plan; Combine the optimized simulation results with the optimized recommended listing plan to determine the optimal product listing plan, and establish an intelligent early warning system to monitor inventory and market dynamics in real time to obtain the initial product selection strategy.

3. The method according to claim 2, characterized in that For the candidate product set, a multi-round simulation environment is constructed using a reinforcement learning algorithm. Virtual agents are introduced to represent the redemption behavior patterns of different types of bank users to simulate the market response under different listing strategies. Parameters are adjusted and optimized based on the user interaction data after each round of simulation to obtain the optimized simulation results, including: Using the reinforcement learning algorithm, a multi-round simulation environment is constructed and processed according to the product listing cycle within a preset time period. Multiple time slices are divided in each round to simulate the changes in market demand in different periods, thus obtaining a multi-round simulation environment. Based on historical redemption behaviors, browsing records, cross-platform behavior trajectories, and social network analysis, a preset redemption behavior model is constructed, and a group of virtual agents are introduced, and the virtual agents are used to optimize in the simulation process to generate a virtual agent group, and the virtual agent group is deployed in a multi-round simulation environment; For each set of candidate listing products, multiple listing strategies are set as input variables, and the multiple listing strategies are configured in multiple rounds of simulation environments to generate a configured simulation environment; In the configured simulation environment, the virtual agent group interacts with the target round's listing strategy according to the preset redemption behavior pattern to simulate actual user redemption behavior, and uses natural language processing technology to perform sentiment analysis on unstructured user comments to generate user interaction data and sentiment tendency scores during the simulation period; Based on the user interaction data and sentiment tendency scores collected after each round of simulation, counterfactual reasoning is used to evaluate the effects of unimplemented listing strategies, and the parameters of the reinforcement learning model are dynamically adjusted and optimized to obtain the optimized simulation parameters.

4. The method according to claim 3, characterized in that: In the configured simulation environment, the virtual agent group interacts with the target round's listing strategy according to the preset redemption behavior pattern, simulating actual user redemption behavior, and uses natural language processing technology to perform sentiment analysis on unstructured user comments to generate user interaction data and sentiment tendency scores during the simulation, including: Using a virtual agent group, deployed in a multi-round simulation environment, the virtual agent group interacts with the target round's listing strategy, the virtual agent selects a commodity according to a preset redemption behavior pattern, simulates the redemption behavior, and generates simulated redemption behavior data; Collect simulated exchange behavior data, record the exchange decisions and results of the virtual agent in each time slice, generate exchange paths and frequency statistics, and obtain preliminary user interaction data; Natural language processing technology is introduced to conduct sentiment analysis on unstructured user comments generated by virtual agents during the simulation process of multiple rounds of simulation environments, identify and quantify users' emotional tendencies, and evaluate user satisfaction in combination with simulated redemption behavior data to generate emotional tendency scores.

5. The method according to any one of claims 1, characterized in that Based on the multi-dimensional user preference map, external market dynamics and social media discussion heat, we use natural language processing technology to perform sentiment analysis on unstructured data, predict the categories of popular redemption products and the potential change paths of popular redemption products, and generate dynamic prediction results, including: By using the node and edge weight information in the multi-dimensional user preference graph, the historical points redemption behavior, browsing history, and cross-platform behavior trajectories of bank users are integrated to obtain the user's interest trend and social influence factor; Collect and analyze external market dynamic information, identify and process external factors that affect user exchange preferences, and obtain external market influencing factors; Monitor the discussion heat on social media, obtain relevant posts, comments and shared content through crawler technology and text mining tools, and generate social media discussion datasets; Using natural language processing technology to perform sentiment analysis on the social media discussion dataset and other unstructured data, quantify the sentiment tendencies of bank users, and obtain sentiment analysis results; A comprehensive prediction model is constructed by combining user interest trends, social influence factors, external market influence factors and sentiment analysis results. The comprehensive prediction model is used to predict the popular redemption commodity categories and the potential change paths of popular redemption commodities to generate dynamic prediction results.

6. The method according to any one of claims 1, characterized in that Implement product listing according to the product listing plan in the initial product selection strategy, collect user interaction data after the product is listed, establish a fast feedback channel, optimize the initial product selection strategy in real time, and generate the best product selection strategy, including: Using the product listing plan in the initial product selection strategy, the products of the points redemption mall are actually arranged and listed to obtain a list of listed products; Based on the list of products on the shelves, the interaction data between bank users and the products on the shelves are collected through various channels to generate a user interaction data set; Establish a fast feedback channel to transmit the user interaction data set to the data analysis platform in real time, so that the user interaction data set is timely and accurate; Analyzing and processing the user interaction data set on a data analysis platform to identify actual preferences and behavior patterns of bank users and obtain a user behavior analysis report; Based on user behavior analysis reports and context-aware computing, we evaluate the effectiveness of the initial product selection strategy, identify areas that need adjustment, and generate strategy adjustment suggestions; According to the strategy adjustment suggestions, the machine learning algorithm is used to optimize the initial product selection strategy to obtain the optimal product selection strategy.

7. The method according to claim 1, characterized in that Utilize the historical points redemption behavior, browsing history, and cross-platform behavior tracks of bank users to build a multi-dimensional user preference map, including: Collect users' historical points redemption behaviors, browsing records, and cross-platform behavior trajectories from multiple channels of the banking system to form a user behavior data set; Preprocess and extract features of user behavior data sets to identify each user's behavior patterns and preference features, and obtain a user behavior feature library; Each user is regarded as a node in the graph, and the edges between nodes represent the association or similarity between users. The weight of the edge is initially set to the default value to reflect the basic degree of association between users and generate the initial user preference graph. According to the points redemption behavior, browsing frequency and dwell time in the user behavior feature library, the user's interest intensity is quantified to obtain the user's interest score; Combined with social network analysis technology, it evaluates the social relationships and influence between users, identifies key opinion leaders and close communities in social networks, and generates social influence factor evaluation results; According to the user interest scores and social influence factor evaluation results, the weights of the edges in the initial user preference map are dynamically adjusted so that the edge weights reflect the interest similarity and social influence between users in real time, and a dynamically updated multi-dimensional user preference map is obtained.

8. An intelligent product selection system for shopping mall products based on bank users' points redemption, characterized in that: include: A construction module is used to build a multi-dimensional user preference map by utilizing the historical points redemption behavior, browsing history, and cross-platform behavior tracks of bank users; The prediction module is used to perform sentiment analysis on unstructured data based on multi-dimensional user preference maps, external market dynamics, and social media discussion heat, combined with natural language processing technology, to predict the categories of popular redemption products and the potential change paths of popular redemption products, and generate dynamic prediction results; A computing module, for generating a customized product recommendation list for each bank user based on the dynamic prediction results, integrating context-aware computing, and computing environmental variables; The simulation module is used to use the product recommendation list, current inventory status and supply chain response time of all bank users, and adopt the reinforcement learning algorithm to simulate the market demand response under different listing strategies, determine the optimal product listing plan, and obtain the initial product selection strategy; The optimization module is used to implement product listing according to the product listing plan in the initial product selection strategy, collect user interaction data after the product is listed, establish a fast feedback channel, optimize the initial product selection strategy in real time, and generate the optimal product selection strategy.

9. A computing device, characterized in that It comprises a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement the intelligent product selection method for listing goods in a mall based on bank users' points redemption as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that: A computer program is stored, and when the computer program is executed by a computer, the intelligent product selection method for listing goods in a mall based on bank users' points redemption is implemented as described in any one of claims 1 to 7.

Images