Product recommendation method and device, computer equipment and storage medium

Abstract

The invention discloses a product recommendation method and device, computer equipment and a storage medium, belongs to the technical field of artificial intelligence, and is used for determining recommended customers adapted to insurance products. Firstly, an advanced natural language processing model is utilized to accurately analyze a user intention, and key semantic information is extracted; secondly, user requirements are accurately mapped to specific industries and business scenes through semantic mapping rules; newest market data and policy information are obtained in real time through a networking search interface, and timeliness and comprehensiveness of the data are guaranteed. Based on a preset report template, the system automatically integrates the upstream and downstream structures and policy dynamics of the industrial chain, and generates a structured and visual market insight report. And finally, extracting high-value customers meeting business requirements from the report through a customer portrait matching algorithm, and forming an accurate recommendation customer set in combination with multi-dimensional feature evaluation. According to the invention, precise product recommendation and high-value customer mining based on natural language input of the user are realized.

Description

Technical Field

[0001] This application belongs to the field of artificial intelligence technology, specifically relating to a product recommendation method, apparatus, computer equipment, and storage medium. Background Technology

[0002] Property insurance product recommendation is a crucial area in the insurance industry, directly impacting an institution's market competitiveness and business growth. Through precise market insights and customer matching, insurance companies can effectively formulate industry-specific strategies and seize opportunities across the value chain. However, current property insurance recommendation schemes have significant limitations in practical application. Many schemes rely on manual analysis or fixed templates, making it difficult to adapt to the rapidly changing market environment and diverse customer needs, especially when dealing with complex industry data, lacking dynamic integration and intelligent matching capabilities. This often leads to sales personnel missing business opportunities due to information lag or insufficient analytical depth when formulating market strategies.

[0003] The core challenge lies in efficiently parsing users' natural language input and combining it with dynamic market information to generate accurate market insights and customer recommendations. First, parsing a user's "one-sentence" question requires the technology to accurately understand business intent, ensuring key information is extracted from brief input. For example, a business person might ask, "What new opportunities are there in the new energy industry?" The system needs to accurately identify keywords such as "new energy" and "opportunities," and connect them to complex backgrounds like policies and the industry chain. Inaccurate parsing may result in reports that deviate from actual needs, impacting decision-making. Second, the dynamic nature of online information retrieval requires the system to quickly filter high-quality, reliable content from massive amounts of data and integrate it into structured market insights. Insufficient search and integration capabilities may lead to fragmented or outdated report content, making it difficult to support accurate business opportunity matching. For example, a branch office might want to develop insurance product recommendations for new energy vehicle companies, but if the system cannot obtain the latest policies or industry chain dynamics in a timely manner, the recommended customer list may be out of touch with the actual market, missing out on high-value customers.

[0004] Therefore, how to build an intelligent system that can accurately analyze users' natural language input, dynamically integrate the latest market information, and generate structured reports has become a key issue in the field of property insurance product recommendation. Summary of the Invention

[0005] The purpose of this application is to propose a product recommendation method, apparatus, computer device, and storage medium to build a system that can accurately parse user natural language input, dynamically integrate the latest market information, and generate structured reports.

[0006] To address the aforementioned technical problems, this application provides a product recommendation method, employing the following technical solution:

[0007] A product recommendation method includes:

[0008] Obtain the natural language text input by the user, and perform semantic analysis of the user intent on the natural language text using a pre-established natural language processing model to obtain a set of intent keywords;

[0009] Based on the set of intent keywords, the system uses preset semantic mapping rules to match industry sectors and business needs, thereby determining the target industries and needs involved in the user's question.

[0010] Starting from the target industry and target needs, real-time market data and policy information are obtained through online search interfaces to obtain the raw dataset;

[0011] For the original dataset, an information filtering algorithm is used to select content that is highly relevant to the set of intent keywords and whose data source is reliable, thus obtaining a refined set of market information.

[0012] Based on a refined market information set, the system integrates upstream and downstream information and policy dynamics of the industry chain through a preset report generation template to generate a structured market insight report.

[0013] From the structured market insight report, a customer profile matching algorithm is used to extract a list of high-value customers that match business needs, resulting in a recommended customer set.

[0014] To address the aforementioned technical problems, this application also provides a product recommendation device, which employs the following technical solution:

[0015] A product recommendation device, comprising:

[0016] The semantic analysis module is used to acquire the natural language text input by the user, and to perform semantic analysis of the user intent on the natural language text through a pre-established natural language processing model to obtain a set of intent keywords.

[0017] The rule matching module is used to match industry fields and business needs based on the set of intent keywords and using preset semantic mapping rules to determine the target industry and target needs involved in the user's question.

[0018] The data acquisition module is used to obtain real-time market data and policy information from the target industry and target needs through the network search interface to obtain the raw dataset.

[0019] The information filtering module is used to filter out content that is highly relevant to the set of intent keywords and whose data source is reliable from the original dataset, thereby obtaining a refined set of market information.

[0020] The report generation module is used to generate structured market insight reports by integrating upstream and downstream information and policy dynamics of the industry chain based on a refined market information set and through preset report generation templates.

[0021] The customer matching module is used to extract a list of high-value customers that match business needs from structured market insight reports using customer profile matching algorithms, and obtain a recommended customer set.

[0022] To address the aforementioned technical problems, this application also provides a computer device that employs the following technical solution:

[0023] A computer device includes a memory and a processor, the memory storing computer-readable instructions, the processor executing the computer-readable instructions to implement the steps of the product recommendation method as described in any of the preceding claims.

[0024] To address the aforementioned technical problems, this application also provides a computer-readable storage medium, employing the technical solution described below:

[0025] A computer-readable storage medium storing computer-readable instructions that, when executed by a processor, implement the steps of the product recommendation method as described in any one of the preceding descriptions.

[0026] Compared with the prior art, the embodiments of this application have the following main advantages:

[0027] This application discloses a product recommendation method, apparatus, computer equipment, and storage medium, belonging to the field of artificial intelligence technology, for determining recommended customers suitable for insurance products. First, it utilizes an advanced natural language processing model to accurately analyze user intent and extract key semantic information. Second, through semantic mapping rules, it precisely maps user needs to specific industries and business scenarios, solving the error problem caused by traditional fuzzy matching. A network search interface obtains the latest market data and policy information in real time, ensuring the timeliness and comprehensiveness of the data. An information filtering algorithm further identifies highly relevant and authoritative data content, improving the reliability and practicality of the information. Based on a preset report template, the system automatically integrates the upstream and downstream structure of the industry chain and policy dynamics to generate a structured and visualized market insight report, helping business personnel quickly understand industry trends and risks / opportunities. Finally, through a customer profile matching algorithm, high-value customers that meet business needs are extracted from the report, and combined with multi-dimensional feature evaluation, a precise set of recommended customers is formed. This application enables precise product recommendations and high-value customer discovery based on user natural language input. It not only significantly improves the efficiency and accuracy of market research and customer screening, but also greatly reduces the manual threshold, realizing the intelligent, automated, and personalized nature of insurance business, and helping enterprises to accurately position the market, enhance competitiveness, and improve customer conversion. Attached Figure Description

[0028] To more clearly illustrate the solutions in this application, the accompanying drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 An exemplary system architecture diagram is shown, in which this application can be applied;

[0030] Figure 2 A flowchart of one embodiment of the product recommendation method according to this application is shown;

[0031] Figure 3 A schematic diagram of one embodiment of the product recommendation device according to this application is shown;

[0032] Figure 4 A schematic diagram of the structure of one embodiment of a computer device according to this application is shown. Detailed Implementation

[0033] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein in the specification of the application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application; the terms "comprising" and "having," and any variations thereof, in the specification, claims, and foregoing drawings of this application, are intended to cover non-exclusive inclusion. The terms "first," "second," etc., in the specification, claims, or foregoing drawings of this application are used to distinguish different objects, not to describe a particular order.

[0034] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0035] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0036] like Figure 1As shown, system architecture 100 may include terminal device 101, network 102, and server 103. Terminal device 101 may be a laptop 1011, tablet 1012, or mobile phone 1013. Network 102 is used as a medium to provide a communication link between terminal device 101 and server 103. Network 102 may include various connection types, such as wired, wireless communication links, or fiber optic cables.

[0037] Users can use terminal device 101 to interact with server 103 via network 102 to receive or send messages, etc. Various communication client applications can be installed on terminal device 101, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social media platform software, etc.

[0038] Terminal device 101 can be various electronic devices with a display screen and support web browsing. In addition to laptops 1011, tablets 1012, or mobile phones 1013, terminal device 101 can also be an e-book reader, an MP3 player (Moving Picture Experts Group Audio Layer III), an MP4 player (Moving Picture Experts Group Audio Layer IV), a laptop computer, and a desktop computer, etc.

[0039] Server 103 can be a server that provides various services, such as a backend server that provides support for the pages displayed on terminal device 101.

[0040] It should be noted that the product recommendation method provided in this application embodiment is generally executed by a server / terminal device, and correspondingly, the product recommendation device is generally set in the server / terminal device.

[0041] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative; the system can have any number of terminal devices, networks, and servers depending on implementation needs.

[0042] Continue to refer to Figure 2 A flowchart of an embodiment of the product recommendation method according to this application is shown. The product recommendation method includes the following steps:

[0043] S201, Obtain the natural language text input by the user, and perform user intent semantic analysis on the natural language text through a pre-established natural language processing model to obtain a set of intent keywords;

[0044] Specifically, a natural language understanding engine based on pre-trained large language models (such as BERT, T5, or GPT) is used to semantically parse the free text input by the user. The model first performs lexical and syntactic analysis to identify core verbs, noun phrases, and modifying structures. Then, an intent classification model (usually based on a multi-label classification network) identifies the user's intent type, such as "industry research," "risk identification," or "insurance product recommendation." Building on this, Named Entity Recognition (NER) technology is further used to extract proper nouns such as region names, industry names, product names, and policy terms from the text. Simultaneously, a contextual attention mechanism is used to determine the semantic relationships between keywords, such as whether "new energy" refers to "industry" or "technology application," or whether "data security" belongs to "risk type" or "regulatory requirement." Keyword extraction uses a Keyword Extraction Network to model explicit and implicit keywords, outputting a structured intent keyword set that includes explicit keywords (existing in the original text) and implicit intent keywords (inferred).

[0045] S202, Based on the set of intent keywords, use preset semantic mapping rules to match industry fields and business needs, and determine the target industry and target needs involved in the user's question;

[0046] Specifically, after obtaining the user's set of intent keywords, this step uses a semantic mapping rule engine to map and match the keywords with a predefined industry standard library (such as the National Economic Industry Classification GB / T 4754) and a business requirement library. This semantic mapping rule library, built by industry experts and a data annotation team, includes lexical synonym expansion, hierarchical relationship mapping (e.g., "intelligent manufacturing" is classified under "advanced manufacturing"), and contextual condition matching rules (e.g., "wind power" in the context of "Jiangsu" corresponds to "offshore wind power" rather than "onshore wind power"). The matching mechanism employs semantic similarity modeling based on graph neural networks, constructing a bidirectional graph between keywords and industry tags, and calculating a context-weighted similarity score. Simultaneously, a multi-classifier is used to identify scenario tags associated with business requirements, such as "risk management," "customer expansion," and "product innovation," thereby determining the industry category and specific business intent of the user input at the semantic level. The matching results will include the target industry ID, industry tags, standard terms, and business scenario tags, along with a confidence score.

[0047] S203, starting from the target industry and target needs, obtains real-time market data and policy information through the network search interface to obtain the raw dataset;

[0048] Specifically, the system utilizes structured industry and demand tags to automatically generate search queries. It then uses an online search interface to access multiple data sources (such as news portals, government websites, industry research institution APIs, and knowledge bases) to obtain real-time data related to the target topic. The search strategy combines Boolean search logic (AND / OR conditions) with a semantic expansion mechanism, automatically embedding synonyms, near-synonyms, and upstream / downstream industry terms related to the keywords. For example, a search for "green finance" can be expanded to include keyword groups such as "investment," "ESG policy," and "carbon asset market." The search interface supports the fusion of structured data sources (such as JSON / XML format policy databases) and unstructured data sources (such as HTML web pages and PDF reports). Leveraging web crawlers and an API scheduling system, it requests data in parallel to acquire the most comprehensive information possible within a set time window. Simultaneously, it uses data collection metadata (such as timestamps, publishing institutions, and content types) to provide credibility criteria for data filtering. The final output is a raw dataset containing the original text content, metadata, and data source tags.

[0049] S204. For the original dataset, an information filtering algorithm is used to select content that is highly relevant to the set of intent keywords and whose data source is reliable, thus obtaining a refined set of market information.

[0050] Specifically, the information filtering module performs multi-level processing on the original dataset to ensure that the retained data content is highly relevant to user intent, has a credible source, and can be used for decision support. The technical implementation includes three core components: relevance assessment, content deduplication, and credibility filtering. First, in the relevance assessment stage, a dual-tower encoder model based on the Transformer architecture is constructed to vectorize the intent keyword set and the original data content, and documents are ranked by semantic matching scores. Second, content deduplication is performed, including redundant paragraph removal based on sentence vector similarity, overlapping citation comparison, and templated sentence recognition. Furthermore, the system constructs a source credibility scoring mechanism, integrating multiple dimensions such as the authoritative level of the information source (e.g., government websites > commercial media > forums and blogs), publication time, information completeness, and verifiability to calculate the final credibility score. Only content with both relevance and credibility scores exceeding set thresholds proceeds to the next step. The final output, a refined market information set, is high-quality text data that has undergone semantic filtering, logical consistency verification, and source security screening, including annotated paragraph titles, content summaries, data sources, and embedded industry tags, among other metadata.

[0051] S205, based on a refined market information set, integrates upstream and downstream information and policy dynamics of the industry chain through a preset report generation template to generate a structured market insight report;

[0052] Specifically, based on a structured and refined set of market information, the system automatically generates market insight reports using a template-driven generation method in collaboration with a language model. First, the system constructs various content structure templates for different industries based on industry classification systems, value chain models, and policy type libraries, such as a five-segment structure: "Industry Chain Structure - Upstream and Downstream Enterprises - Core Policies - Risk Points - Insurance Opportunities." Then, the language model performs semantic clustering and sentence rewriting on the refined data, ensuring that content from different sources is integrated into narrative paragraphs with a consistent language style and logical coherence. In particular, the model categorizes and organizes content based on extracted industry chain keywords (such as "raw material processing," "core manufacturing," and "end-user applications"), and analyzes policy dynamics in conjunction with a time dimension, revealing time trends and potential changes. Furthermore, a syntax-controlled generation template is introduced to guide the model in generating content with industry terminology and professional formatting. The report output supports multiple formats such as JSON, Word, and PDF, and can be used for API integration or front-end display, ensuring that the structured information is easy for both machines to read and humans to understand.

[0053] S206. From the structured market insight report, a customer profile matching algorithm is used to extract a list of high-value customers that match business needs, resulting in a recommended customer set.

[0054] Specifically, after generating the market insight report, the system identifies potential customers from the enterprise information database using a customer profiling matching algorithm, based on the target industry, upstream and downstream supply chains, and key needs. The customer profiling system consists of multi-dimensional features, including basic enterprise information (registered location, registered capital, annual revenue), business scope, main business, industries involved, publicly reported activities, related upstream and downstream enterprises, and existing insurance coverage. First, a matching vector is constructed using high-frequency keywords, industry tags, and policy signals extracted from the report. This vector is then matched against the structured vector of the enterprise information to calculate semantic similarity. Second, the system uses rule-based matching and machine learning models (such as LightGBM or BERT-based matching models) to determine the degree of matching between the enterprise and business needs. Finally, based on a comprehensive score considering industry priority, enterprise size, activity level, and recent investment and financing activities, high-value customers with high intent alignment are selected, and a recommended customer list is output. This list supports further access to business personnel relationship graphs, establishing connection paths with existing teams to assist in lead follow-up.

[0055] Furthermore, based on the set of intent keywords, pre-defined semantic mapping rules are used to match industry sectors and business needs, determining the target industry and target needs involved in the user's question. This process specifically includes:

[0056] Each keyword in the intent keyword set is compared with each rule entry in the semantic mapping rule base to find the industry domain tag and business requirement tag that best match each keyword.

[0057] The matched industry tags and business requirement tags are combined to form a preliminary judgment on the industry and needs involved in the user's question;

[0058] By combining users' historical question records and preference settings, the initial judgment was revised and optimized, and the specific target industry and target needs involved in the user's question were finally determined.

[0059] In this embodiment, the system compares the set of intent keywords with the rule entries in the semantic mapping rule base one by one. Based on an embedded semantic vector model (such as Word2Vec or BERT Embedding), the system calculates the semantic similarity between each keyword and the rule entry, and selects the industry tag and business demand tag with the highest similarity as candidates. A context-aware mechanism is introduced to avoid incorrect matching due to semantic ambiguity. For example, "carbon" in the context of "carbon trading" should match the environmental protection industry, not the materials industry. The matching results employ a multi-tag scoring mechanism and set a confidence threshold to ensure coverage of the multi-dimensional aspects of user intent. The matched tags are then combined to form multiple sets of industry-demand candidate pairs. The system further calls upon the user's historical question logs and preference configurations (such as regional preferences and industry priority) to sort and optimize the candidate results through logistic regression or attention weighting mechanisms, improving personalized accuracy. For example, if a user has repeatedly asked about "green manufacturing," even if the current input vaguely mentions "energy-saving equipment," the system will prioritize associating it with the "clean energy equipment manufacturing" field. Finally, a high-confidence industry-demand combination is output.

[0060] By following the steps above, semantic ambiguity can be effectively avoided, the accuracy of industry identification and the refinement of user needs understanding can be achieved, and the relevance of recommended content and user satisfaction can be improved.

[0061] Furthermore, the process of comparing each keyword in the intent keyword set with each rule entry in the semantic mapping rule base to find the industry domain tag and business requirement tag that best matches each keyword includes:

[0062] Based on the preset similarity calculation rules, the similarity between each keyword and the industry domain tag and business requirement tag in each rule entry is calculated respectively;

[0063] Select the industry-specific tags and business-needs tags with the highest similarity as the most matching tags for the current keywords;

[0064] If multiple keywords correspond to the same industry domain tag or business demand tag, the weight of the industry domain tag or business demand tag will be increased.

[0065] By iteratively processing all keywords in the intent keyword set, an industry domain tag set and a business requirement tag set corresponding to the intent keyword set are formed.

[0066] In this embodiment, the system first generates a context-aware semantic vector (e.g., using BERT or RoBERTa models) for each keyword in the intent keyword set. Then, it calculates the similarity between this vector and the semantic vectors of the industry domain labels and business requirement labels corresponding to each rule in the semantic mapping rule base. Similarity calculation can employ cosine similarity, Euclidean distance, or a weighted matching model based on an attention mechanism. For each keyword, the system selects the industry domain label and business requirement label with the highest similarity from all labels as the most matching label for that keyword and records their matching score. If multiple keywords are mapped to the same label, the label's weight is automatically accumulated based on the matching frequency and similarity score, forming a confidence index for the candidate label. To enhance robustness, the system also detects semantic redundancy and label overfitting, ensuring a balance between label set diversity and expression accuracy. Finally, through iterative processing of the entire intent keyword set, two structured sets are generated: an industry domain label set and a business requirement label set. This set, based on label name, weight score, and source keywords, serves as the core input for industry identification and content generation and can be extended to support user preference weight adjustment, custom filtering, and other strategies.

[0067] Through the above steps, high-precision semantic alignment and weight modeling of user-input intent keywords can be achieved, improving the consistency and accuracy of industry and demand identification.

[0068] Furthermore, starting from the target industry and target needs, the steps to obtain the raw dataset by acquiring real-time market data and policy information through online search interfaces specifically include:

[0069] Based on the industry domain tag set and the business requirement tag set, construct a combination of search keywords tailored to user needs;

[0070] By utilizing online search interfaces, combined search keywords can be submitted to various information sources for retrieval.

[0071] The search results from different information sources are collected and integrated to form the original dataset, which contains multi-dimensional information such as market data, policy documents, and industry dynamics.

[0072] In this embodiment, the system constructs multi-level search keyword combinations based on industry domain tag sets and business demand tag sets, employing a keyword expansion strategy. These combinations include core tag words, synonymous expanded words, upstream and downstream related terms, and policy-related words. For example, when the industry tag is "intelligent manufacturing" and the business demand is "risk assessment," the system automatically generates keyword groups such as "intelligent manufacturing risk supervision policy," "industrial internet security vulnerabilities," and "intelligent equipment market trends." Keyword combination construction utilizes knowledge graph-based entity expansion and hierarchical relationship extraction technology to ensure coverage of multiple dimensions within the industry. After filtering, the generated keyword groups are submitted to multiple information sources through the system's preset online search interface, including government official websites (such as the Ministry of Industry and Information Technology), industry association platforms, financial news media, data portal websites, and enterprise information platforms. The retrieval process employs asynchronous concurrent scheduling, combined with anti-crawling strategies and data collection frequency limiting control, to ensure crawling efficiency and stability. All returned results will be initially categorized according to document type (such as news, policy documents, market analysis reports, etc.), and metadata (source, time, publisher, summary, etc.) will be extracted, labeled, and archived. Finally, they will be uniformly integrated to form a raw dataset containing multi-dimensional content such as market data, policy information, and industry dynamics.

[0073] Through the above steps, the system can quickly build a keyword retrieval scheme with high coverage and high relevance, ensuring that the original data is comprehensive, real-time and structurally diverse, and significantly improving the depth and breadth of data collection.

[0074] Furthermore, based on the refined market information set, the steps of integrating upstream and downstream information and policy dynamics of the industry chain through a preset report generation template to generate a structured market insight report specifically include:

[0075] The information in the refined market information collection is categorized and sorted according to the preset report generation template;

[0076] We extract key information and data from critical nodes in the upstream and downstream of the industry chain, and conduct in-depth analysis in conjunction with policy dynamics to form industry insights.

[0077] We combine industry insights with market demands and present them visually through charts, videos, and text.

[0078] Embedding visual content into the report generation template generates structured market insight reports.

[0079] In this embodiment, the system first categorizes the content in the refined market information set according to a preset report generation template. The template typically includes modules such as "Policy Guidance," "Industry Chain Structure," "Upstream and Downstream Enterprise Dynamics," "Industry Hotspots," "Potential Risks," and "Insurance Intervention Points." Information classification employs an automatic paragraph classification algorithm based on tags and semantic embedding, while also considering the timeliness of content publication, information density, and the authority of the original source for ranking. Next, the system delves into key nodes in the upstream and downstream of the industry chain, extracting information on core links such as "raw material suppliers," "core manufacturing links," and "distribution terminals" through entity recognition and upstream / downstream relationship recognition models. It then logically deduces and identifies trends in their operational status and policy influencing factors to form industry insights. During this process, the system also aggregates and analyzes relevant policy information to identify the policy support and regulatory direction for a specific link in the industry chain. Subsequently, the industry insights are matched with user business needs, generating visual charts (such as upstream / downstream flowcharts and trend analysis charts), dynamic diagrams (such as policy evolution timelines), and accompanying textual interpretations for key content, improving user comprehension efficiency. Ultimately, these visualizations are automatically embedded into preset templates, outputting structured market insight reports with a unified format, clear structure, and professional content. The reports support multiple output formats, including Word, PDF, and web pages, making them easy to view and edit.

[0080] Through the above steps, the system can transform scattered and heterogeneous market information into highly structured industry reports, enhancing the reports' professionalism, readability, and visual presentation capabilities.

[0081] Furthermore, the steps involved in extracting a list of high-value customers matching business needs from the structured market insight report using a customer profiling matching algorithm to obtain a recommended customer set include:

[0082] Based on industry insights, market demands, and characteristics of high-value customers in the structured market insight report, a customer profile model is constructed.

[0083] The customer profile model is compared with the preset customer database to filter out potential customers who meet business needs and have a high degree of feature matching.

[0084] For the selected potential customers, a multi-dimensional evaluation algorithm is used to screen out high-value customers. The multi-dimensional evaluation includes credit score, historical transaction record, and industry status.

[0085] The selected high-value customers are sorted by priority to form a recommended customer list, and a recommended customer set is built based on the recommended customer list.

[0086] In this embodiment, the system first analyzes key elements in the structured market insight report, such as industry insights (e.g., emerging sectors, policy-supported areas), market demands (e.g., financing, risk transfer, compliance assurance), and the behavioral and attribute characteristics of potential customers, to construct a multi-dimensional customer profile model. This model includes static attributes (e.g., company size, registered capital, industry, geographical distribution) and dynamic behavioral data (e.g., industry chain role, public cooperation records, policy response behavior, news popularity), and combines machine learning algorithms (e.g., K-Means clustering analysis, XGBoost classification model) to model the similarity of potential customers. The system then performs batch comparisons of this profile model with a pre-set enterprise database to retrieve a list of potential customers with high feature matching and those whose industry chain position aligns with the report. Next, a multi-dimensional evaluation mechanism is introduced to perform credit scoring (based on publicly available business, judicial, and financial data), transaction history analysis (e.g., past cooperation records with insurance institutions), and industry influence analysis (e.g., whether they are leading enterprises or core suppliers), assigning each customer a quantitative value score. Finally, based on the comprehensive scoring results, the customers are prioritized and automatically generated into a recommended customer list, which is then output in a set format. This list can be integrated with sales or CRM systems for business follow-up and marketing task distribution.

[0087] For example, in a consultation regarding "green manufacturing risk assessment," the user input "insurance needs of new energy battery manufacturers" as natural language. The system, through semantic intent parsing, extracted keywords such as "new energy," "battery," "production," and "risk assessment," and determined the target industry as "clean energy equipment manufacturing" and the demand scenario as "risk management in the production process" using semantic mapping rules. Subsequently, the platform constructed extended search terms including "clean energy equipment production risk supervision," "battery manufacturing safety compliance," and "new energy process quality standards." Through online search, it integrated the latest government safety regulatory policies, industry research reports, and news updates, forming a raw dataset containing policy white papers, market size data, and key equipment accident cases. After filtering out non-authoritative sources and low-relevance content, the information filtering algorithm yields high-value policy interpretation paragraphs, industry chain maps, and typical accident analyses. Based on this, the report generation template, using a five-segment framework of "industry chain structure—upstream materials—core manufacturing—downstream applications—policy evolution," automatically arranges and visualizes equipment supplier distribution maps, accident frequency line graphs, and policy timelines, generating a market insight report rich in text and charts. Finally, using "key equipment suppliers," "high-risk production links," and "policy subsidy recipients" from the report as profiling factors, the system filters out eligible potential customers from the enterprise database and performs multi-dimensional evaluation and ranking based on credit scores, cooperation records, and industry position, outputting a clearly prioritized list of high-value customers.

[0088] Through the above steps, the system can accurately identify high-value customers that are highly aligned with business objectives, thereby improving the relevance, convertibility, and follow-up efficiency of customer recommendations.

[0089] Furthermore, for the selected potential customers, a multi-dimensional evaluation algorithm is used to screen out high-value customers. This process specifically includes:

[0090] Calculate the credit score for each potential customer based on a pre-set credit scoring model.

[0091] By combining the historical transaction records of each potential customer, we analyze the past cooperation data between each potential customer and the insurance company to obtain a past cooperation score. The past cooperation data includes the frequency of cooperation, transaction amount, and contract execution status.

[0092] By referencing the industry position information of each potential client, we assess each potential client's market share, brand influence, and technological innovation capabilities in their respective industries to obtain an industry position score.

[0093] The overall value score of each potential customer is calculated using a weighted average algorithm, taking into account credit score, past cooperation score, and industry status score.

[0094] High-value customers are selected based on their overall value score.

[0095] In this embodiment, the system first conducts a risk assessment for each potential customer based on a pre-set credit scoring model, combined with publicly available business registration information, legal risk databases, financial statements, and other multi-source data, quantifying the credit score. This model typically integrates multiple indicators, such as corporate solvency, historical default records, and financial health, using machine learning algorithms to dynamically adjust weights to improve scoring accuracy. Subsequently, the system deeply analyzes the potential customer's historical transaction data with the insurance company, examining the frequency of cooperation, transaction amount, and contract performance, quantifying past cooperation scores. This step reflects not only the stability of customer cooperation but also potential commercial value and risk controllability. Next, considering the customer's position within the industry, the system comprehensively considers market share data, brand influence evaluation, and technological innovation capabilities, using third-party market research reports and patent data to score the industry position score. Finally, the system weights the credit score, past cooperation score, and industry position score according to pre-set weights to form a comprehensive value score. The weighting configuration can be flexibly adjusted according to different insurance product strategies; for example, higher-risk emerging industries may place greater emphasis on credit scores. Customers whose overall score is higher than the threshold are selected as high-value customers. The system supports dynamic adjustment of the threshold to adapt to market changes, thereby achieving automated and precise customer selection.

[0096] A property insurance company is conducting high-value customer screening for "intelligent manufacturing equipment suppliers." The system first collects financial reports, business registration information, and legal risk data of target customers based on a credit scoring model. Combining this with debt repayment ability, historical litigation records, and tax compliance, a pre-trained random forest model is used to calculate a credit score for a potential customer, A. The model's weights are: debt repayment ability 40%, historical litigation risk 30%, and tax compliance 30%. Customer A scores 85 points (out of 100) in debt repayment ability, 90 points (low historical litigation risk), and 70 points (moderate tax compliance). The overall credit score is calculated as: 85 × 0.4 + 90 × 0.3 + 70 × 0.3 = 34 + 27 + 21 = 82 points.

[0097] Next, the system evaluates Client A's past cooperation with the insurance company. Analysis of transaction records over the past three years reveals that Client A's cumulative transaction amount reached 12 million yuan, with an average cooperation frequency of 3 times per year, and no contract breaches. To quantify the cooperation score, a maximum score of 15 million yuan was set for the cooperation amount, a maximum score of 5 times per year for the frequency, and a penalty of 20 points for each breach. Client A's cooperation amount score is 1200 / 1500 × 100 = 80 points, the cooperation frequency score is 3 / 5 × 100 = 60 points, and the breach penalty is 0. The weighted calculation of the cooperation score is 60% for amount, 30% for frequency, and 10% for breach, resulting in a comprehensive cooperation score of: 80 × 0.6 + 60 × 0.3 + 0 × 0.1 = 48 + 18 + 0 = 66 points.

[0098] Subsequently, the industry position score was completed through market research data and patent information analysis. Client A's market share accounts for approximately 15% of the segment, corresponding to a score of 75 (out of 100); brand influence, based on an industry questionnaire, received 70 points; and technological innovation capability, based on the quantity and quality of patents applied for in the past 5 years, scored 80 points. The weighted average of the industry position score was 50% for market share, 30% for brand influence, and 20% for technological innovation, calculated as follows: 75 × 0.5 + 70 × 0.3 + 80 × 0.2 = 37.5 + 21 + 16 = 74.5 points.

[0099] Finally, the system comprehensively evaluates the three scores. Assuming the credit score has a weight of 50%, the cooperation score has a weight of 30%, and the industry status score has a weight of 20%, then Customer A's comprehensive value score is: 82 × 0.5 + 66 × 0.3 + 74.5 × 0.2 = 41 + 19.8 + 14.9 = 75.7 points. This score is higher than the system's preset threshold of 70 points, therefore Customer A is judged as a high-value customer and included in the priority recommendation list.

[0100] Through the above steps, the system scientifically selects customer groups that meet business needs and have outstanding value based on multi-dimensional quantitative indicators, significantly improving the accuracy of customer recommendations and business conversion efficiency.

[0101] In the above embodiments, this application discloses a product recommendation method, belonging to the field of artificial intelligence technology, for determining suitable customers for insurance products. First, an advanced natural language processing model is used to accurately analyze user intent and extract key semantic information. Second, semantic mapping rules are used to precisely map user needs to specific industries and business scenarios, solving the error problem caused by traditional fuzzy matching. A network search interface obtains the latest market data and policy information in real time, ensuring the timeliness and comprehensiveness of the data. Information filtering algorithms further identify highly relevant and authoritative data content, improving the reliability and practicality of the information. Based on preset report templates, the system automatically integrates the upstream and downstream structure of the industry chain and policy dynamics to generate structured and visualized market insight reports, helping business personnel quickly understand industry trends and risks / opportunities. Finally, a customer profile matching algorithm extracts high-value customers that meet business needs from the reports, and combined with multi-dimensional feature evaluation, forms a precise set of recommended customers. This application enables precise product recommendations and high-value customer discovery based on user natural language input. It not only significantly improves the efficiency and accuracy of market research and customer screening, but also greatly reduces the manual threshold, realizing the intelligent, automated, and personalized nature of insurance business, and helping enterprises to accurately position the market, enhance competitiveness, and improve customer conversion.

[0102] In this embodiment, the product recommendation method operates on an electronic device (e.g., Figure 1 The server shown can receive instructions or acquire data via wired or wireless connection. It should be noted that the aforementioned wireless connection methods may include, but are not limited to, 3G / 4G connections, WiFi connections, Bluetooth connections, WiMAX connections, Zigbee connections, UWB (ultra-wideband) connections, and other currently known or future wireless connection methods.

[0103] It should be emphasized that, to further ensure the privacy and security of the aforementioned user intent information, this user intent information can also be stored in a node of a blockchain.

[0104] The blockchain referred to in this application is a novel application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and encryption algorithms. Essentially, a blockchain is a decentralized database, a chain of data blocks linked together using cryptographic methods. Each data block contains information about a batch of network transactions, used to verify the validity of the information (anti-counterfeiting) and generate the next block. A blockchain can include an underlying blockchain platform, a platform product service layer, and an application service layer.

[0105] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0106] Foundational technologies for artificial intelligence generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies mainly encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0107] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by instructing related hardware through computer-readable instructions. These computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the methods described above. The aforementioned storage medium can be a non-volatile storage medium such as a magnetic disk, optical disk, or read-only memory (ROM), or random access memory (RAM).

[0108] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0109] Further reference Figure 3 As a response to the above Figure 2 To implement the method shown, this application provides an embodiment of a product recommendation device, which is similar to... Figure 2 Corresponding to the method embodiments shown, this device can be specifically applied to various electronic devices.

[0110] like Figure 3 As shown, the product recommendation device 300 described in this embodiment includes:

[0111] The semantic analysis module 301 is used to acquire the natural language text input by the user, and to perform user intent semantic analysis on the natural language text through a pre-established natural language processing model to obtain a set of intent keywords.

[0112] The rule matching module 302 is used to match industry fields and business needs based on the set of intent keywords and using preset semantic mapping rules to determine the target industry and target needs involved in the user's question.

[0113] The data acquisition module 303 is used to obtain real-time market data and policy information from the target industry and target needs through the network search interface to obtain the raw dataset.

[0114] The information filtering module 304 is used to filter out content that is highly relevant to the set of intent keywords and whose data source is reliable from the original dataset using an information filtering algorithm, thereby obtaining a refined set of market information.

[0115] The report creation module 305 is used to generate a structured market insight report by integrating upstream and downstream information and policy dynamics of the industry chain based on a refined market information set and through a preset report generation template.

[0116] The customer matching module 306 is used to extract a list of high-value customers that match business needs from the structured market insight report using a customer profile matching algorithm, and obtain a recommended customer set.

[0117] Furthermore, the rule matching module 302 specifically includes:

[0118] The rule matching unit is used to compare each keyword in the intent keyword set with each rule entry in the semantic mapping rule base one by one to find the industry domain tag and business requirement tag that best match each keyword.

[0119] The tag combination unit is used to combine the matched industry domain tags and business requirement tags to form a preliminary judgment on the industry and needs involved in the user's question;

[0120] The target optimization unit is used to revise and optimize the initial judgment by combining the user's historical question records and preference settings, and finally determine the specific target industry and target needs involved in the user's question.

[0121] Furthermore, the rule matching unit specifically includes:

[0122] The similarity calculation subunit is used to calculate the similarity between each keyword and the industry domain tag and business requirement tag in each rule entry according to the preset similarity calculation rules.

[0123] The tag matching subunit is used to select the industry domain tags and business demand tags with the highest similarity as the most matching tags for the current keyword;

[0124] The tag-weighted subunit is used to increase the weight of the industry domain tag or business demand tag if multiple keywords correspond to the same industry domain tag or business demand tag.

[0125] The tag aggregation sub-unit is used to iteratively process all keywords in the intent keyword set to form an industry domain tag set and a business requirement tag set corresponding to the intent keyword set.

[0126] Furthermore, the data acquisition module 303 specifically includes:

[0127] The search keyword unit is used to construct a combination of search keywords tailored to user needs based on industry domain tag sets and business requirement tag sets;

[0128] The big data retrieval unit is used to submit combinations of search keywords to various information sources for retrieval via the network search interface;

[0129] The retrieval and aggregation unit is used to collect and integrate retrieval results from different information sources to form a raw dataset, which contains multi-dimensional information such as market data, policy documents, and industry dynamics.

[0130] Furthermore, the report creation module 305 specifically includes:

[0131] The classification and sorting unit is used to classify and sort the information in the refined market information set according to the preset report generation template.

[0132] The in-depth analysis unit is used to extract key information and data from critical nodes in the upstream and downstream of the industry chain, and to conduct in-depth analysis in conjunction with policy dynamics to form industry insights.

[0133] The visualization unit is used to combine industry insights with market demands and present them visually through charts, videos, and text.

[0134] The report creation unit is used to embed visual content into report generation templates to generate structured market insight reports.

[0135] Furthermore, the customer matching module 306 specifically includes:

[0136] The customer profiling unit is used to build customer profile models based on industry insights, market demands, and characteristics of high-value customers in the structured market insight report.

[0137] The profile comparison unit is used to compare the customer profile model with the preset customer database to filter out potential customers who meet business needs and have a high degree of feature matching.

[0138] The evaluation and screening unit is used to select high-value customers from the potential customers using a multi-dimensional evaluation algorithm. The multi-dimensional evaluation includes credit score, historical transaction record, and industry status.

[0139] The priority sorting unit is used to sort the selected high-value customers according to their priority to form a recommended customer list, and to build a recommended customer set based on the recommended customer list.

[0140] Furthermore, the evaluation and screening units specifically include:

[0141] The credit scoring subunit is used to calculate the credit score of each potential customer based on a preset credit scoring model.

[0142] The Past Cooperation Scoring Subunit is used to analyze the past cooperation data between each potential customer and the insurance company by combining the historical transaction records of each potential customer, and to obtain the past cooperation score. The past cooperation data includes cooperation frequency, transaction amount, and contract execution status.

[0143] The industry position scoring sub-unit is used to refer to the industry position information of each potential customer, to evaluate each potential customer’s market share, brand influence and technological innovation capabilities in the industry, and to obtain an industry position score.

[0144] The scoring weighted sub-unit is used to combine credit score, past cooperation score and industry status score, and uses a weighted average algorithm to calculate the comprehensive value score of each potential customer.

[0145] The evaluation and screening sub-unit is used to screen high-value customers based on their overall value score.

[0146] In the above embodiments, this application discloses a product recommendation device, belonging to the field of artificial intelligence technology, for determining recommended customers suitable for insurance products. First, an advanced natural language processing model is used to accurately analyze user intent and extract key semantic information. Second, semantic mapping rules are used to accurately map user needs to specific industries and business scenarios, solving the error problem caused by traditional fuzzy matching. A network search interface obtains the latest market data and policy information in real time, ensuring the timeliness and comprehensiveness of the data. Information filtering algorithms further identify highly relevant and authoritative data content, improving the reliability and practicality of the information. Based on preset report templates, the system automatically integrates the upstream and downstream structure of the industry chain and policy dynamics to generate structured and visualized market insight reports, helping business personnel quickly understand industry trends and risks / opportunities. Finally, a customer profile matching algorithm extracts high-value customers that meet business needs from the reports, and combined with multi-dimensional feature evaluation, forms a precise set of recommended customers. This application enables precise product recommendations and high-value customer discovery based on user natural language input. It not only significantly improves the efficiency and accuracy of market research and customer screening, but also greatly reduces the manual threshold, realizing the intelligent, automated, and personalized nature of insurance business, and helping enterprises to accurately position the market, enhance competitiveness, and improve customer conversion.

[0147] To address the aforementioned technical problems, embodiments of this application also provide a computer device. Please refer to [link / reference needed]. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0148] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected via a system bus. It should be noted that only the computer device 4 with memory 41, processor 42, and network interface 43 is shown in the figure; however, it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented alternatively. Those skilled in the art will understand that the computer device described here is a device capable of automatically performing numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes, but is not limited to, microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0149] The computer device can be a desktop computer, laptop, handheld computer, or cloud server, etc. The computer device can interact with the user via a keyboard, mouse, remote control, touchpad, or voice control.

[0150] The memory 41 includes at least one type of readable storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 41 may be an internal storage unit of the computer device 4, such as the hard disk or memory of the computer device 4. In other embodiments, the memory 41 may also be an external storage device of the computer device 4, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the computer device 4. Of course, the memory 41 may also include both the internal storage unit and its external storage device of the computer device 4. In this embodiment, the memory 41 is typically used to store the operating system and various application software installed on the computer device 4, such as computer-readable instructions for product recommendation methods. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or will be output.

[0151] In some embodiments, the processor 42 may be a central processing unit (CPU), controller, microcontroller, microprocessor, or other data processing chip. The processor 42 is typically used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to execute computer-readable instructions stored in the memory 41 or to process data, such as executing computer-readable instructions for the product recommendation method.

[0152] The network interface 43 may include a wireless network interface or a wired network interface, which is typically used to establish communication connections between the computer device 4 and other electronic devices.

[0153] This application also provides an embodiment, namely, a computer device including a memory and a processor. The memory stores computer-readable instructions, and the processor, when executing the computer-readable instructions, implements the steps of the product recommendation method described above, that is, implements:

[0154] A product recommendation method includes:

[0155] Obtain the natural language text input by the user, and perform semantic analysis of the user intent on the natural language text using a pre-established natural language processing model to obtain a set of intent keywords;

[0156] Based on the set of intent keywords, the system uses preset semantic mapping rules to match industry sectors and business needs, thereby determining the target industries and needs involved in the user's question.

[0157] Starting from the target industry and target needs, real-time market data and policy information are obtained through online search interfaces to obtain the raw dataset;

[0158] For the original dataset, an information filtering algorithm is used to select content that is highly relevant to the set of intent keywords and whose data source is reliable, thus obtaining a refined set of market information.

[0159] Based on a refined market information set, the system integrates upstream and downstream information and policy dynamics of the industry chain through a preset report generation template to generate a structured market insight report.

[0160] From the structured market insight report, a customer profile matching algorithm is used to extract a list of high-value customers that match business needs, resulting in a recommended customer set.

[0161] This application also provides another embodiment, namely, a computer-readable storage medium storing computer-readable instructions that can be executed by at least one processor to cause the at least one processor to perform the steps of the product recommendation method described above, i.e., to implement:

[0162] A product recommendation method includes:

[0163] Obtain the natural language text input by the user, and perform semantic analysis of the user intent on the natural language text using a pre-established natural language processing model to obtain a set of intent keywords;

[0164] Based on the set of intent keywords, the system uses preset semantic mapping rules to match industry sectors and business needs, thereby determining the target industries and needs involved in the user's question.

[0165] Starting from the target industry and target needs, real-time market data and policy information are obtained through online search interfaces to obtain the raw dataset;

[0166] For the original dataset, an information filtering algorithm is used to select content that is highly relevant to the set of intent keywords and whose data source is reliable, thus obtaining a refined set of market information.

[0167] Based on a refined market information set, the system integrates upstream and downstream information and policy dynamics of the industry chain through a preset report generation template to generate a structured market insight report.

[0168] From the structured market insight report, a customer profile matching algorithm is used to extract a list of high-value customers that match business needs, resulting in a recommended customer set.

[0169] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0170] This application can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics devices, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0171] It should be noted that the software tools or components not belonging to this company that appear in the various embodiments of this application are merely illustrative examples and do not represent actual use.

[0172] Obviously, the embodiments described above are only some embodiments of this application, not all embodiments. The accompanying drawings show preferred embodiments of this application, but do not limit the patent scope of this application. This application can be implemented in many different forms; rather, the purpose of providing these embodiments is to provide a more thorough and comprehensive understanding of the disclosure of this application. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing specific embodiments, or make equivalent substitutions for some of the technical features. Any equivalent structures made using the content of this application's specification and drawings, directly or indirectly applied to other related technical fields, are similarly within the scope of patent protection of this application.

Claims

1. A product recommendation method characterized by, The method comprises the following steps: acquiring natural language text input by a user, performing user intent semantic analysis on the natural language text by a pre-established natural language processing model to obtain an intent keyword set; matching an industry field and a business demand by using a preset semantic mapping rule according to the intent keyword set to determine a target industry and a target demand involved in the user question; obtaining an original data set by acquiring real-time market data and policy information through a network search interface starting from the target industry and the target demand; obtaining a refined market information set by screening out content highly related to the intent keyword set and having reliable data sources from the original data set by using an information filtering algorithm; generating a structured market insight report by integrating upstream and downstream information of an industry chain and policy dynamics through a preset report generation template according to the refined market information set; obtaining a recommended customer set by extracting a high-value customer list matched with the business demand from the structured market insight report by using a customer portrait matching algorithm.

2. The product recommendation method according to claim 1, wherein The step of matching an industry field and a business demand by using a preset semantic mapping rule according to the intent keyword set to determine a target industry and a target demand involved in the user question comprises the following steps: comparing each keyword in the intent keyword set with each rule item in a semantic mapping rule library one by one to find an industry field label and a business demand label most matched with each keyword; combining the matched industry field label and the business demand label to form a preliminary judgment of the industry and the demand involved in the user question; combining user historical question records and preference settings to correct and optimize the preliminary judgment to finally determine the specific target industry and the target demand involved in the user question.

3. The product recommendation method according to claim 2, wherein The step of comparing each keyword in the intent keyword set with each rule item in the semantic mapping rule library one by one to find an industry field label and a business demand label most matched with each keyword comprises the following steps: calculating the similarity of each keyword with an industry field label and a business demand label in each rule item according to a preset similarity calculation rule; selecting an industry field label and a business demand label with the highest similarity as the most matched label of the current keyword; if there are multiple keywords corresponding to the same industry field label or business demand label, increasing the weight of the industry field label or the business demand label; processing all keywords in the intent keyword set by iteration to form an industry field label set and a business demand label set corresponding to the intent keyword set.

4. The product recommendation method according to claim 3, wherein The step of acquiring real-time market data and policy information through a network search interface starting from the target industry and the target demand to obtain an original data set comprises the following steps: constructing a search keyword combination for user demand according to the industry field label set and the business demand label set; submitting the search keyword combination to each information source for retrieval by using the network search interface; Collect and integrate search results from different information sources to form the original data set, wherein the original data set contains multi-dimensional information of market data, policy documents and industry trends.

5. The product recommendation method according to claim 1, wherein The step of generating a structured market insight report according to the refined market information set and through a preset report generation template to integrate upstream and downstream information of the industry chain and policy trends, specifically includes: Classifying and sorting the information in the refined market information set according to the preset report generation template; For the key nodes of the upstream and downstream of the industry chain, extract key information and data, and conduct in-depth analysis combined with policy trends to form industry insight points; Combine industry insight points with market demand, and visualize through charts, videos and text; Embed the visualized content into the report generation template to generate the structured market insight report.

6. The product recommendation method according to claim 5, wherein The step of extracting a high-value customer list matched with the business demand from the structured market insight report using a customer portrait matching algorithm to obtain a recommended customer set, specifically includes: According to the industry insight points, market demand and high-value customer characteristics in the structured market insight report, construct a customer portrait model; Compare the customer portrait model with a preset customer database to filter out potential customers that meet business demand and have high feature matching degree; For the filtered potential customers, use a multi-dimensional evaluation algorithm to filter out high-value customers, wherein the multi-dimensional evaluation includes credit score, historical transaction record and industry status; Sort the filtered high-value customers according to priority to form a recommended customer list, and build the recommended customer set based on the recommended customer list.

7. The product recommendation method according to claim 6, wherein The step of filtering out high-value customers from the filtered potential customers using a multi-dimensional evaluation algorithm, specifically includes: According to a preset credit score model, calculate the credit score of each potential customer; Combine the historical transaction records of each potential customer to analyze the past cooperation data between each potential customer and the insurance company to obtain a past cooperation score, wherein the past cooperation data includes cooperation frequency, transaction amount and contract execution; Referring to the industry status information of each potential customer, evaluate the market share, brand influence and technological innovation capability of each potential customer in the industry to obtain an industry status score; Integrate the credit score, past cooperation score and industry status score to calculate the comprehensive value score of each potential customer using a weighted average algorithm; According to the high and low of the comprehensive value score of each potential customer, filter out the high-value customers.

8. A product recommendation device characterized by comprising: It includes: A semantic analysis module for obtaining a natural language text input by a user, performing user intent semantic analysis on the natural language text through a pre-established natural language processing model, and obtaining an intent keyword set; A rule matching module for matching an industry field and a business demand using a preset semantic mapping rule according to the intent keyword set to determine a target industry and a target demand involved in the user's question; A data acquisition module for obtaining real-time market data and policy information through a network search interface from the target industry and the target demand to obtain an original data set; An information filtering module is configured to filter, for the original data set, content highly related to the set of intent keywords and reliable in data source by using an information filtering algorithm to obtain a refined market information set; A report making module is configured to integrate upstream and downstream information of an industry chain and policy dynamics by using a preset report generation template to generate a structured market insight report according to the refined market information set; A customer matching module is configured to extract a list of high-value customers matched with the business demand from the structured market insight report by using a customer portrait matching algorithm to obtain a recommended customer set.

9. A computer device, comprising: A product recommendation method comprises a memory and a processor, the memory stores computer readable instructions, and the processor executes the computer readable instructions to implement the steps of the product recommendation method according to any one of claims 1 to 7.

10. A computer readable storage medium characterized by The computer readable storage medium stores computer readable instructions, and the computer readable instructions are executed by the processor to implement the steps of the product recommendation method according to any one of claims 1 to 7.

Images