Product recommendation method and device, electronic equipment and storage medium

Abstract

The invention discloses a product recommendation method and device, electronic equipment and a storage medium, and relates to the technical field of data processing, basic information and demand preferences of a user are obtained, a multi-dimensional dynamic feature vector is constructed, the vector can dynamically capture multiple demands of the user in different stages of a life cycle, and the user experience is improved. The method does not depend on a fixed preset rule or single-dimension data; and generating a recommendation request based on the dynamic vector, and inputting the recommendation request into a recommendation algorithm module to screen and sort products, so that recommendation logic can closely fit real-time demand change of a user, and limitation of static analysis of a traditional model is avoided, and therefore, deviation between a recommendation result and an actual demand of the user caused by an existing preset rule or a traditional machine learning model can be solved; the technical problems that the recommendation accuracy and the user satisfaction are affected are solved, and the technical effects of improving the recommendation accuracy of the personal pension insurance products, effectively matching the actual demands of the users in different stages and improving the satisfaction of the users to recommendation services are achieved.

Description

Technical Field

[0001] This disclosure relates to the field of data processing technology, and more particularly to a recommended method and apparatus for a product, an electronic device, and a storage medium. Background Technology

[0002] With the continuous development of artificial intelligence technology, its application in the financial services sector is becoming increasingly profound, particularly demonstrating significant potential in insurance product recommendation. Personal pension insurance, as an important component at the intersection of finance and social security, is widely used in scenarios such as user lifecycle management, risk protection planning, and asset allocation optimization. Among related technologies, a preliminary intelligent recommendation system has been constructed through the collaborative operation of rule engines, basic algorithm models, and user behavior analysis.

[0003] Currently, directly using preset rules or traditional machine learning models to implement the above-mentioned intelligent recommendation system may lead to deviations between the recommendation results and the user's actual needs, and may fail to adapt to changes in user preferences throughout the user's lifecycle, thereby affecting the accuracy of the recommendation and user satisfaction. Summary of the Invention

[0004] This disclosure provides a product recommendation method, apparatus, electronic device, and storage medium. Its main purpose is to address the problem that recommendation results deviate from actual user needs and fail to adapt to changes in user preferences throughout the user lifecycle, thereby affecting the accuracy of recommendations and user satisfaction.

[0005] According to a first aspect of this disclosure, a method for recommending a product is provided, comprising: Obtain basic information and user preferences from user input, and construct a multi-dimensional dynamic feature vector of the user based on the basic information and user preferences; Recommendation requests are generated based on the aforementioned multi-dimensional dynamic feature vectors; The recommendation request is input into the recommendation algorithm module, which filters and sorts the products to generate a personalized recommendation list for the user. The user is displayed according to the personalized recommendation list.

[0006] Optionally, after displaying the information to the user based on the personalized recommendation list, the method further includes: Collect user feedback on the personalized recommendation list; Based on the feedback information, the recommendation algorithm module is adjusted and optimized in real time.

[0007] Optionally, obtaining basic information and user preferences input by the user, and constructing a multi-dimensional dynamic feature vector of the user based on the basic information and user preferences, includes: By analyzing user voice or text input using natural language processing technology, semantic features of users regarding pension eligibility age, eligibility method, risk preference, and protection needs can be extracted. A federated learning framework is used to process user data for privacy protection and merge it with local user data to generate a multi-dimensional dynamic feature vector that includes static attributes, dynamic behaviors, and external environment.

[0008] Optionally, generating a recommendation request based on the multi-dimensional dynamic feature vector includes: By using deep reinforcement learning models to perform real-time reasoning on user profiles, we can identify the current lifecycle stage of users and their potential implicit needs. The weights of each feature in the recommendation request are dynamically adjusted based on the user's current interaction context and historical behavior patterns to optimize the adaptability of the recommendation strategy.

[0009] Optionally, the step of inputting the recommendation request into the recommendation algorithm module, filtering and sorting products, and generating the user's personalized recommendation list includes: The recommendation request is input into a reinforcement learning model based on Markov decision process modeling; wherein, the reward function of the reinforcement learning model includes three dimensions: user satisfaction, product suitability, and long-term benefits; An attention mechanism is used to weight key features in user profiles to improve the responsiveness of recommendation ranking to core user needs.

[0010] Optionally, the method further includes: A three-dimensional feedback matrix of users, products, and environment is constructed based on user feedback information, and the recommendation strategy is updated in real time through online learning algorithms to form a closed-loop optimized recommendation system.

[0011] According to a second aspect of this disclosure, a product recommendation device is provided, comprising: The acquisition unit is used to acquire basic information and user preferences input by the user, and to construct a multi-dimensional dynamic feature vector of the user based on the basic information and user preferences. A generation unit is used to generate recommendation requests based on the multi-dimensional dynamic feature vector; The generation unit is also used to input the recommendation request into the recommendation algorithm module, filter and sort the products, and generate the user's personalized recommendation list; The display unit is used to display information to the user based on the personalized recommendation list.

[0012] Optionally, the device further includes: The collection unit is used to collect feedback information from the user on the personalized recommendation list after the display unit performs the display on the user according to the personalized recommendation list; The adjustment unit is used to adjust and optimize the recommendation algorithm module in real time based on the feedback information.

[0013] Optionally, the acquisition unit is further configured to: By analyzing user voice or text input using natural language processing technology, semantic features of users regarding pension eligibility age, eligibility method, risk preference, and protection needs can be extracted. A federated learning framework is used to process user data for privacy protection and merge it with local user data to generate a multi-dimensional dynamic feature vector that includes static attributes, dynamic behaviors, and external environment.

[0014] Optionally, the generation unit is further configured to: By using deep reinforcement learning models to perform real-time reasoning on user profiles, we can identify the current lifecycle stage of users and their potential implicit needs. The weights of each feature in the recommendation request are dynamically adjusted based on the user's current interaction context and historical behavior patterns to optimize the adaptability of the recommendation strategy.

[0015] Optionally, the generation unit is further configured to: The recommendation request is input into a reinforcement learning model based on Markov decision process modeling; wherein, the reward function of the reinforcement learning model includes three dimensions: user satisfaction, product suitability, and long-term benefits; An attention mechanism is used to weight key features in user profiles to improve the responsiveness of recommendation ranking to core user needs.

[0016] Optionally, the device further includes: The building unit is used to construct a three-dimensional feedback matrix of users, products, and environment based on user feedback information, and to update the recommendation strategy in real time through online learning algorithms to form a closed-loop optimized recommendation system.

[0017] According to a third aspect of this disclosure, an electronic device is provided, comprising: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method described in the first aspect above.

[0018] According to a fourth aspect of this disclosure, a non-transitory computer-readable storage medium is provided storing computer instructions, wherein the computer instructions are configured to cause the computer to perform the method described in the first aspect above.

[0019] According to a fifth aspect of this disclosure, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the method described in the first aspect above.

[0020] This application, by first acquiring basic user information and preferences and constructing a multi-dimensional dynamic feature vector, can dynamically capture the diverse needs of users at different stages of their life cycle, rather than relying on fixed preset rules or single-dimensional data. Then, based on this dynamic vector, a recommendation request is generated and input into the recommendation algorithm module to filter and rank products. This allows the recommendation logic to closely align with real-time changes in user needs, avoiding the limitations of static analysis in traditional models. Therefore, it can solve the technical problem of deviations between recommendation results and actual user needs caused by existing preset rules or traditional machine learning models, affecting recommendation accuracy and user satisfaction. This achieves the technical effect of improving the accuracy of personal pension insurance product recommendations, effectively matching the actual needs of users at different stages, and increasing user satisfaction with the recommendation service.

[0021] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent from the following description. Attached Figure Description

[0022] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein: Figure 1 A schematic flowchart illustrating a product recommendation method provided in an embodiment of this disclosure; Figure 2 A schematic diagram of the structure of a product recommendation device provided in an embodiment of this disclosure; Figure 3 A schematic diagram of the structure of a product recommendation device provided in an embodiment of this disclosure; Figure 4 A schematic block diagram of an example electronic device provided for embodiments of this disclosure. Detailed Implementation

[0023] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0024] The following description, with reference to the accompanying drawings, outlines recommended methods, apparatuses, electronic devices, and storage media for products according to embodiments of this disclosure.

[0025] Figure 1 This is a schematic flowchart illustrating a product recommendation method provided in an embodiment of this disclosure.

[0026] like Figure 1 As shown, the method includes the following steps: Step 101: Obtain the user's basic information and needs preferences, and construct the user's multi-dimensional dynamic feature vector based on the basic information and needs preferences; Basic information encompasses static attributes such as age, gender, occupation, and income, which are crucial for characterizing a user's fundamental situation. Demand preferences include personalized needs directly related to retirement planning, such as the user's desired retirement age, risk tolerance, and preferred retirement insurance product payment methods. Users can provide this information through various interactive methods, including text input, voice input, or a graphical interface, ensuring convenient and accurate information delivery. After acquiring this information, the system performs structured processing, integrating the scattered basic information and demand preference data to construct a multi-dimensional dynamic feature vector for the user. The multi-dimensional aspect refers to the vector simultaneously containing both static basic attributes and dynamic demand preferences, comprehensively reflecting the user's individual characteristics. The dynamic aspect means that when the user subsequently adjusts their demand preferences or supplements basic information, the feature vector can be updated accordingly, maintaining consistency with the user's current situation and providing a reliable user characteristic basis for the system to conduct accurate retirement insurance product matching.

[0027] Step 102: Generate a recommendation request based on the multi-dimensional dynamic feature vector; The AI ​​Agent module first performs deep analysis of the acquired multi-dimensional dynamic feature vectors, accurately identifying key information such as the user's static basic attributes (age, occupation, income) and dynamic needs and preferences (expected pension receiving age, risk tolerance), fully understanding the user's core demands in choosing pension insurance products. Next, the AI ​​Agent module combines the user's current interaction scenario—whether the recommendation request is triggered while actively browsing a pension insurance product page or explicitly searching for related products—to integrate the complete information from the analyzed multi-dimensional dynamic feature vectors, the extracted core user needs and preferences, and the current recommendation scenario information, forming a structured recommendation request. This recommendation request clearly and comprehensively conveys relevant user information, providing accurate and clear input for the subsequent recommendation algorithm module to filter and rank pension insurance products. This ensures that the recommendation algorithm module can accurately match pension insurance products that meet the user's needs based on this request, laying a solid foundation for personalized recommendations.

[0028] Step 103: Input the recommendation request into the recommendation algorithm module to filter and sort products, and generate a personalized recommendation list for the user; The recommendation algorithm module first receives the recommendation request, and then analyzes the key information contained in the request, such as the user's multi-dimensional dynamic feature vector, core needs and preferences, and recommendation scenarios. It clarifies the user's static attributes, such as age, occupation, and income; dynamic demands, such as expected retirement age and risk tolerance; and the scenario type that triggers the recommendation, such as browsing or searching.

[0029] The recommendation algorithm module calls appropriate algorithm models, which may include collaborative filtering, content-based recommendation, or deep learning recommendation algorithms, selecting the optimal algorithm combination based on the current recommendation scenario and user characteristics. In the screening stage, the algorithm matches user characteristics with the core attributes of pension insurance products. These core attributes include rate of return, risk level, and payout method. For example, for users with low risk tolerance, low-risk products are prioritized, while high-risk products are excluded; for users who wish to receive their pension early, products with earlier payout start ages are prioritized.

[0030] After the initial screening, the algorithm proceeds to the sorting stage. This sorting is based on a weighted allocation of user preferences and needs. If a user prioritizes product yield, the filtered products are sorted from highest to lowest yield. If a user values ​​withdrawal flexibility, the flexibility of the withdrawal method is the primary sorting criterion. Simultaneously, the algorithm is optimized to incorporate latent user needs. For example, deep learning algorithms are used to uncover implicit preferences from users' historical behavior, further adjusting the sorting results to ultimately generate a personalized list of pension insurance product recommendations tailored to each user's individual needs. This provides a clear basis for product selection when presented to users later.

[0031] Step 104: Display the information to the user based on the personalized recommendation list.

[0032] The user interaction module first obtains the personalized recommendation list generated in step 103, and integrates and processes the core information of each pension insurance product in the list. This core information includes key content related to user decision-making, such as product name, rate of return, risk level, pension payment method, and starting age for receiving pension benefits. It ensures that the information is complete and clearly expressed, and avoids affecting user understanding due to missing or obscure information.

[0033] The user interaction module selects an appropriate display method based on the user's previous interaction habits. If the user previously entered information through a graphical interface, the system will use the graphical interface as the primary display medium, presenting a personalized recommendation list in sorted order. Each product will have an independent display item, with key product information clearly listed in separate sections within each item, allowing users to easily compare the differences between different products. If the user previously used voice input, the system will support voice interaction-assisted display in addition to the graphical interface. When the user issues a voice query command, the system can read out the key information of the specified product in real time, catering to different user preferences.

[0034] During the demonstration, the user interaction module will also set up convenient operation entry points for each recommended product. Users can click on the corresponding product item to view more comprehensive information such as detailed product terms and fees, ensuring that users can fully obtain the information they need during the demonstration, and prepare for subsequent product selection or feedback, effectively realizing the effective transmission between the recommendation results and users.

[0035] In some embodiments, after displaying a recommendation to the user based on the personalized recommendation list, the method further includes: Collect user feedback on the personalized recommendation list; Based on the feedback information, the recommendation algorithm module is adjusted and optimized in real time.

[0036] The user interaction module proactively collects user feedback on the personalized recommendation list. This feedback includes both behavioral and proactive feedback. Behavioral feedback includes data such as click-through rates for different products in the recommendation list, whether users clicked to view product details, and whether they completed a purchase based on the recommended products. Proactive feedback includes user satisfaction ratings for recommended products, descriptions of preferences for certain product attributes, or explanations of reasons for disapproval of the recommendations submitted through the interactive interface. Users can conveniently submit proactive feedback through text input, graphical interface selection, and other methods, ensuring the comprehensiveness and timeliness of the feedback information.

[0037] The feedback optimization module receives feedback from the user interaction module and performs structured analysis and data mining on this information. For example, if a user has a very low click rate on high-yield but high-risk products in the recommendation list and submits feedback indicating a greater concern for fund security, the feedback optimization module will analyze and conclude that the current recommendation algorithm module is underweighting the user's risk tolerance feature. Based on this analysis, the feedback optimization module will adjust and optimize the recommendation algorithm module in real time. This includes adjusting the weight ratio of each dimension of the user's multi-dimensional dynamic feature vector in the algorithm model, such as increasing the weight of the risk tolerance feature, or updating the matching rules when the algorithm selects products, such as strengthening the priority of low-risk products in the selection process. This ensures that the recommendation algorithm module can more accurately meet the user's actual needs when processing the user's recommendation requests, forming a closed loop of recommendation-feedback-optimization, and continuously improving the accuracy of recommendations.

[0038] In some embodiments, obtaining basic information and user preferences input by the user, and constructing a multi-dimensional dynamic feature vector of the user based on the basic information and user preferences, includes: By analyzing user voice or text input using natural language processing technology, semantic features of users regarding pension eligibility age, eligibility method, risk preference, and protection needs can be extracted. A federated learning framework is used to process user data for privacy protection and merge it with local user data to generate a multi-dimensional dynamic feature vector that includes static attributes, dynamic behaviors, and external environment.

[0039] The system utilizes integrated natural language processing (NLP) technology to process user input. This technology can accurately analyze a user's voice or text content. For example, if a user voices, "I want to start receiving my pension at 55, monthly, with moderate risk, and with medical insurance," or texts, "I want to start receiving my pension at 60, annually, with moderate risk tolerance, and a focus on ensuring basic living standards after retirement," the NLP technology will perform semantic analysis, entity recognition, and intent extraction on this content. From this, it will extract key semantic features related to the user's pension receiving age (55 or 60), receiving method (monthly or annually), risk preference (low or moderate risk), and insurance needs (medical insurance or basic living security), ensuring accurate capture of the user's personalized needs.

[0040] The system employs a federated learning framework to perform privacy-preserving processing on the extracted user data and user-related data obtained from other channels. The federated learning framework enables data value mining through distributed training and parameter sharing without directly accessing raw user data, avoiding user privacy leaks. For example, when integrating user career income or historical financial behavior data, there is no need for centralized data storage; model parameters are only passed after data processing on local nodes. Subsequently, the privacy-preserving data is fused with locally stored user static attributes such as age, gender, and occupation, and dynamic behaviors such as past browsing records of pension insurance products. This results in a multi-dimensional dynamic feature vector containing static attributes, dynamic behaviors, and external environmental information such as current changes in the basic interest rate of the pension insurance market. This provides comprehensive and secure user feature support for subsequent recommendation request generation and product matching.

[0041] In some embodiments, generating a recommendation request based on the multi-dimensional dynamic feature vector includes: By using deep reinforcement learning models to perform real-time reasoning on user profiles, we can identify the current lifecycle stage of users and their potential implicit needs. The weights of each feature in the recommendation request are dynamically adjusted based on the user's current interaction context and historical behavior patterns to optimize the adaptability of the recommendation strategy.

[0042] The AI ​​Agent module utilizes a deep reinforcement learning model. This model takes the user's multi-dimensional dynamic feature vectors as its core input and performs real-time inference operations on the user profile. During inference, the model deeply analyzes the static attributes contained in the feature vectors, such as age, occupation, and income, as well as dynamic preferences such as risk tolerance and pension expectations. This allows for accurate identification of the user's current life cycle stage. For example, a 28-year-old young professional corresponds to the initial stage of retirement planning, while a 58-year-old nearing retirement corresponds to the stage of perfecting retirement security. Simultaneously, the model can also uncover potential implicit needs that users haven't explicitly expressed. For instance, while younger users may not mention these needs, their job stability and income growth trends suggest a demand for flexible payment methods, while those nearing retirement may implicitly have a higher requirement for stable pension payments.

[0043] The AI ​​Agent module dynamically adjusts the weights of various features in recommendation requests by combining the user's current interaction context and historical behavior patterns. If a user triggers a recommendation by actively searching for "low-risk pension insurance," the interaction context indicates a clear need, and the weight of the risk preference feature in the recommendation request will be increased. If a user has repeatedly clicked on "monthly payment" type products in their historical behavior, it indicates a clear preference for payment methods, and the weight of the payment method feature will be increased accordingly. Through this dynamic adjustment of weights, the recommendation request is ensured to accurately reflect the user's current core needs, optimizing the adaptability of the recommendation strategy to different interaction scenarios and user behavior patterns, and providing more relevant guidance for the subsequent recommendation algorithm module to select and rank pension insurance products.

[0044] In some embodiments, the step of inputting the recommendation request into the recommendation algorithm module, filtering and sorting products, and generating the user's personalized recommendation list includes: The recommendation request is input into a reinforcement learning model based on Markov decision process modeling; wherein, the reward function of the reinforcement learning model includes three dimensions: user satisfaction, product suitability, and long-term benefits; An attention mechanism is used to weight key features in user profiles to improve the responsiveness of recommendation ranking to core user needs.

[0045] Markov Decision Processes (MDPs) model the dynamic relationship between "state-action-reward" in recommendation scenarios, accurately capturing the dynamic changes in user needs and product matching during the pension insurance recommendation process. This aligns with the characteristic that pension products must consistently meet user needs. The reward function of this reinforcement learning model comprises three core dimensions: user satisfaction (referencing historical feedback data from similar users on similar products, such as click-through rates and post-use ratings); product fit (comparing the degree of fit between the user's multi-dimensional dynamic feature vectors in the recommendation request and the attributes of the pension insurance product, such as the match between the user's occupational income and the product's contribution amount, and the match between the user's expected retirement age and the product's retirement start time); and long-term benefit (evaluating the long-term pension security effect of the product, such as the accumulated pension amount and the stability of benefits during the retirement period). Through the comprehensive calculation of these three dimensions, a scientific quantitative standard is provided for product selection. Next, the recommendation algorithm module uses an attention mechanism to weight key features in the user profile. The attention mechanism can automatically identify the core features in the user profile that affect the recommendation results. For example, when a user emphasizes "wanting flexible pension payment methods", the mechanism will increase the weight of the "payment method" feature in the user's needs and preferences. During the recommendation ranking stage, products with more flexible payment methods will be given higher ranking priority, thereby improving the responsiveness of the recommendation ranking to the user's core needs. Finally, after screening and ranking, a list of pension insurance products that meet the user's personalized needs is generated.

[0046] In some embodiments, the method further includes: A three-dimensional feedback matrix of users, products, and environment is constructed based on user feedback information, and the recommendation strategy is updated in real time through online learning algorithms to form a closed-loop optimized recommendation system.

[0047] The feedback optimization module collects user feedback on the personalized recommendation list. This information includes user behavior feedback such as product click-through rate, detail view rate, and purchase conversion rate, as well as proactive feedback such as satisfaction rating, description of preferences for product attributes, and reasons for not accepting recommendations. It also records the environmental information when the feedback occurs, including current interest rate changes in the pension insurance market, product policy adjustments, and the interactive environment that triggered the user feedback, such as browsing time and device type.

[0048] The feedback optimization module constructs a three-dimensional feedback matrix based on this information, encompassing users, products, and the environment. The user dimension includes core information such as the user's multi-dimensional dynamic feature vectors, including age, risk tolerance, and expected age for claiming benefits. The product dimension covers key attributes of products in the recommendation list, such as rate of return, risk level, and claiming method. The environment dimension incorporates market and interaction environment data at the time of feedback. Each cell in the matrix corresponds to a specific user's feedback on a specific product within a specific environment, forming a structured set of feedback data.

[0049] The feedback optimization module utilizes an online learning algorithm to perform real-time analysis and processing of the three-dimensional feedback matrix. This online learning algorithm does not rely on offline batch data training and can capture real-time trends in the feedback data within the matrix. For example, if the matrix shows that a certain type of medium-risk, annually paid product has seen a significant increase in purchase conversion rates among multiple users aged 35-45 in an environment of rising market interest rates, the algorithm will quickly identify this correlation and update the recommendation strategy of the recommendation algorithm module in real time. This could involve adjusting the weight of this type of product in the recommendation ranking for the corresponding user group, or optimizing the environmental factor matching rules during product selection.

[0050] User feedback is transformed into effective data support through a three-dimensional matrix. Driven by online learning algorithms, the recommendation strategy iterates in real time, ultimately forming a closed-loop optimization recommendation system of "user feedback - matrix construction - algorithm update - strategy optimization - new recommendation - new feedback". This continuously improves the system's adaptability to user needs and environmental changes, ensuring dynamic optimization of recommendation results.

[0051] Corresponding to the product recommendation method described above, this invention also proposes a product recommendation device. Since the device embodiments of this invention correspond to the method embodiments described above, details not disclosed in the device embodiments can be referred to in the method embodiments described above, and will not be repeated here.

[0052] Figure 2This is a schematic diagram of the structure of a product recommendation device provided in an embodiment of this disclosure, as shown below. Figure 2 As shown, it includes: Acquisition unit 21 is used to acquire basic information and demand preferences input by the user, and construct a multi-dimensional dynamic feature vector of the user based on the basic information and demand preferences; Generation unit 22 is used to generate recommendation requests based on the multi-dimensional dynamic feature vector; The generation unit 22 is also used to input the recommendation request into the recommendation algorithm module, filter and sort the products, and generate the user's personalized recommendation list; Display unit 23 is used to display information to the user based on the personalized recommendation list.

[0053] Furthermore, in one possible implementation of the embodiments of this disclosure, such as Figure 3 As shown, the device further includes: The collection unit 24 is used to collect feedback information from the user on the personalized recommendation list after the display unit 23 performs the display to the user according to the personalized recommendation list; The adjustment unit 25 is used to adjust and optimize the recommendation algorithm module in real time based on the feedback information.

[0054] Furthermore, in one possible implementation of this disclosure, the acquisition unit 21 is further configured to: By analyzing user voice or text input using natural language processing technology, semantic features of users regarding pension eligibility age, eligibility method, risk preference, and protection needs can be extracted. A federated learning framework is used to process user data for privacy protection and merge it with local user data to generate a multi-dimensional dynamic feature vector that includes static attributes, dynamic behaviors, and external environment.

[0055] Furthermore, in one possible implementation of this disclosure embodiment, the generation unit 22 is further configured to: By using deep reinforcement learning models to perform real-time reasoning on user profiles, we can identify the current lifecycle stage of users and their potential implicit needs. The weights of each feature in the recommendation request are dynamically adjusted based on the user's current interaction context and historical behavior patterns to optimize the adaptability of the recommendation strategy.

[0056] Furthermore, in one possible implementation of this disclosure embodiment, the generation unit 22 is further configured to: The recommendation request is input into a reinforcement learning model based on Markov decision process modeling; wherein, the reward function of the reinforcement learning model includes three dimensions: user satisfaction, product suitability, and long-term benefits; An attention mechanism is used to weight key features in user profiles to improve the responsiveness of recommendation ranking to core user needs.

[0057] Furthermore, in one possible implementation of the embodiments of this disclosure, such as Figure 3 As shown, the device further includes: Building unit 26 is used to construct a three-dimensional feedback matrix of users, products, and environment based on user feedback information, and to update the recommendation strategy in real time through an online learning algorithm to form a closed-loop optimized recommendation system.

[0058] It should be noted that the foregoing explanation of the method embodiments also applies to the apparatus of the embodiments of this disclosure, and the principle is the same. Therefore, the embodiments of this disclosure are not limited thereto.

[0059] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0060] Figure 4 A schematic block diagram of an example electronic device 400 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0061] like Figure 4 As shown, device 400 includes a computing unit 401, which can perform various appropriate actions and processes based on a computer program stored in ROM (Read-Only Memory) 402 or a computer program loaded from storage unit 408 into RAM (Random Access Memory) 403. RAM 403 may also store various programs and data required for the operation of device 400. The computing unit 401, ROM 402, and RAM 403 are interconnected via bus 404. I / O (Input / Output) interface 405 is also connected to bus 404.

[0062] Multiple components in device 400 are connected to I / O interface 405, including: input unit 406, such as keyboard, mouse, etc.; output unit 407, such as various types of monitors, speakers, etc.; storage unit 408, such as disk, optical disk, etc.; and communication unit 409, such as network card, modem, wireless transceiver, etc. Communication unit 409 allows device 400 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0063] The computing unit 401 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 401 include, but are not limited to, CPUs (Central Processing Units), GPUs (Graphics Processing Units), various special-purpose AI (Artificial Intelligence) computing chips, various computing units running machine learning model algorithms, DSPs (Digital Signal Processors), and any suitable processor, controller, microcontroller, etc. The computing unit 401 performs the various methods and processes described above, such as the recommended method for a product. For example, in some embodiments, the recommended method for a product may be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 408. In some embodiments, part or all of the computer program may be loaded and / or installed on device 400 via ROM 402 and / or communication unit 409. When the computer program is loaded into RAM 403 and executed by the computing unit 401, one or more steps of the methods described above may be performed. Alternatively, in other embodiments, the computing unit 401 may be configured to perform the recommended method of the aforementioned product by any other suitable means (e.g., by means of firmware).

[0064] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, FPGAs (Field Programmable Gate Arrays), ASICs (Application-Specific Integrated Circuits), ASSPs (Application-Specific Standard Products), SOCs (System-on-Chips), CPLDs (Complex Programmable Logic Devices), computer hardware, firmware, software, and / or combinations thereof. These various implementations may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0065] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0066] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, RAM, ROM, EPROM (Electrically Programmable Read-Only Memory) or flash memory, optical fiber, CD-ROM (Compact Disc Read-Only Memory), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0067] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (Cathode-Ray Tube) or LCD (Liquid Crystal Display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0068] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include LANs (Local Area Networks), WANs (Wide Area Networks), the Internet, and blockchain networks.

[0069] Computer systems can include clients and servers. Clients and servers are generally geographically separated and typically interact via communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. A server can be a cloud server, also known as a cloud computing server or cloud host, a hosting product within the cloud computing service system that addresses the shortcomings of traditional physical hosts and VPS (Virtual Private Server) services, such as high management difficulty and weak business scalability. Servers can also be servers for distributed systems or servers incorporating blockchain technology.

[0070] It's important to note that artificial intelligence (AI) is the study of enabling computers to simulate certain human thought processes and intelligent behaviors (such as learning, reasoning, thinking, and planning). It encompasses both hardware and software technologies. AI hardware technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, and big data processing. AI software technologies primarily include computer vision, speech recognition, natural language processing, machine learning / deep learning, big data processing, and knowledge graph technologies.

[0071] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0072] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A product recommendation method characterized by, The method comprises: obtaining basic information and demand preferences input by a user, and constructing a multi-dimensional dynamic feature vector of the user based on the basic information and the demand preferences; generating a recommendation request based on the multi-dimensional dynamic feature vector; inputting the recommendation request into a recommendation algorithm module, screening and sorting products, and generating a personalized recommendation list for the user; performing display to the user according to the personalized recommendation list.

2. The method of claim 1, wherein, After performing display to the user according to the personalized recommendation list, the method further comprises: collecting feedback information of the user on the personalized recommendation list; based on the feedback information, real-time adjusting and optimizing the recommendation algorithm module.

3. The method of claim 1, wherein, The obtaining basic information and demand preferences input by a user, and constructing a multi-dimensional dynamic feature vector of the user based on the basic information and the demand preferences comprises: analyzing user voice or text input through natural language processing technology, and extracting semantic features of the user on pension receiving age, receiving mode, risk preference and protection demand; adopting a federated learning framework to process user data for privacy protection, and fusing with local user data to generate a multi-dimensional dynamic feature vector containing static attributes, dynamic behaviors and external environment.

4. The method of claim 1, wherein, The generating a recommendation request based on the multi-dimensional dynamic feature vector comprises: real-time reasoning of a user portrait through a deep reinforcement learning model, identifying a life cycle stage currently being in by the user and potential implied demand; according to a current interaction context and historical behavior pattern of the user, dynamically adjusting weights of each feature in the recommendation request to optimize adaptability of a recommendation strategy.

5. The method of claim 1, wherein, The inputting the recommendation request into a recommendation algorithm module, screening and sorting products, and generating a personalized recommendation list for the user comprises: inputting the recommendation request into a reinforcement learning model modeled based on a Markov decision process; wherein a reward function of the reinforcement learning model contains three dimensions of user satisfaction, product adaptation degree and long-term benefit; adopting an attention mechanism to weight process key features in the user portrait, to improve response ability to core demand of the user in recommendation sorting.

6. The method of claim 1, wherein, The method further comprises: constructing a three-dimensional feedback matrix of user, product and environment based on user feedback information, and real-time updating a recommendation strategy through an online learning algorithm to form a closed-loop optimized recommendation system.

7. A product recommendation apparatus characterized by comprising: The method comprises: an obtaining unit, configured to obtain basic information and demand preferences input by a user, and construct a multi-dimensional dynamic feature vector of the user based on the basic information and the demand preferences; a generating unit, configured to generate a recommendation request based on the multi-dimensional dynamic feature vector; the generating unit is further configured to input the recommendation request into a recommendation algorithm module, screen and sort products, and generate a personalized recommendation list for the user; a display unit, configured to perform display to the user according to the personalized recommendation list.

8. An electronic device, comprising: The method comprises: at least one processor; and a memory connected in communication with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method in any one of claims 1-6.

9. A non-transitory computer-readable storage medium having stored thereon computer instructions, wherein, The computer instructions are for causing the computer to perform the method according to any one of claims 1-6.

10. A computer program product, characterised in that, A computer program comprising instructions which, when executed by a processor, implement the method according to any one of claims 1-6.

Images