Artificial intelligence-based strategy generation method and apparatus, computer device, and medium

Abstract

The application belongs to the technical field of artificial intelligence, and relates to a strategy generation method and device based on artificial intelligence, computer equipment and a storage medium, which comprises the following steps: receiving an input strategy generation request; performing intent recognition on input data to obtain corresponding intent information; performing knowledge base retrieval on the intent information to obtain corresponding retrieval results; performing label recommendation processing on the retrieval results to obtain corresponding label recommendation information; performing goods generation processing on the intent information, the retrieval results and the label recommendation information to obtain corresponding target goods; performing strategy generation processing based on the intent information, the retrieval results, the label recommendation information and the target goods to obtain corresponding target strategies; and performing output processing on the target strategies. The application can be applied to a strategy generation scene in the field of financial technology, and effectively improves the processing efficiency and accuracy of strategy generation.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology and can be applied to the financial technology field, particularly to artificial intelligence-based strategy generation methods, devices, computer equipment, and storage media. Background Technology

[0002] In the process of enterprise operation and management, refined operation, experience accumulation, and strategic configuration are key elements for enhancing enterprise competitiveness and achieving sustainable development. However, enterprises still face significant challenges in these areas, with the problem of extensive strategies and a lack of personalized, refined operation being particularly prominent. Specifically, traditional strategy systems exhibit a clear "group-oriented" characteristic in key dimensions such as user reach and resource matching. This characteristic creates a structural conflict with the current trend of continuously upgrading individualized needs. As the market environment becomes increasingly complex and consumer demands become increasingly diversified, users have higher and higher requirements for the personalization of products and services, expecting enterprises to provide precise services and recommendations based on their unique needs and preferences. However, traditional strategy systems struggle to meet this demand and cannot provide tailored operational solutions for different users.

[0003] Existing strategy generation methods generally tend towards uniformity. Currently, enterprises primarily employ rule-based strategy generation. On one hand, manually generated strategies heavily rely on the experience and judgment of professionals. However, ordinary staff, limited by their data interpretation abilities and algorithmic comprehension thresholds, struggle to extract valuable information from massive amounts of data to build accurate customer preference identification systems. They often can only formulate strategies based on superficial, basic information, unable to deeply analyze complex user needs. On the other hand, while rule-based strategy generation offers a degree of objectivity and stability, the rules are often fixed and simple, making it difficult to adapt to constantly changing market environments and user demands. This uniform strategy generation approach makes it difficult for enterprises to develop flexible and precise strategies, failing to meet the personalized needs of diverse users.

[0004] This problem is particularly evident in the scenario of recommending car insurance products. In the process of recommending car insurance products, operations personnel typically rely solely on basic vehicle type classifications or regional tags for resource allocation. For example, users of the same car model in the same region are grouped together and recommended the same car insurance products to them. However, this simplistic classification method ignores the numerous differences between users and fails to incorporate multi-dimensional characteristics such as claims history and coverage preferences into the strategy. Even users with the same car model and location may have vastly different claims histories and coverage preferences. Some users may have good driving habits, fewer claims, and relatively lower coverage requirements; while others may have higher driving risks, more claims, and a desire for higher coverage. Because operations personnel fail to fully consider these multi-dimensional characteristics, the generated strategies lack precision and cannot provide users with car insurance product recommendations that truly meet their needs. This not only reduces the efficiency of strategy generation, requiring operations personnel to spend a significant amount of time and effort on strategy adjustments and optimization, but also compromises the accuracy of the strategies, impacting user satisfaction and the company's market competitiveness.

[0005] Therefore, existing technologies have significant shortcomings in strategy generation and cannot meet the growing demand for personalized operations from enterprises. Thus, there is an urgent need for a new technology solution that can improve the processing efficiency and accuracy of strategy generation. Summary of the Invention

[0006] The purpose of this application is to propose a strategy generation method, apparatus, computer device, and storage medium based on artificial intelligence, so as to solve the technical problem of low processing efficiency and accuracy of existing strategy generation methods.

[0007] Firstly, an artificial intelligence-based strategy generation method is provided, including: Receive a policy generation request; wherein the policy generation request carries input data corresponding to the user; The input data is extracted from the policy generation request, and the input data is subjected to intent recognition based on a preset intent recognition agent to obtain the corresponding intent information; The intent information is retrieved from the knowledge base based on a preset dual search engine to obtain the corresponding search results; The search results are processed by a pre-defined tag recommendation agent to obtain corresponding tag recommendation information. Based on a pre-defined intelligent agent for generating goods, the agent processes the intent information, the search results, and the tag recommendation information to generate the corresponding target goods. Based on the intent information, the search results, the tag recommendation information, and the target product, a strategy generation process is performed to obtain the corresponding target strategy. The target strategy is then processed for output.

[0008] Secondly, an artificial intelligence-based strategy generation device is provided, comprising: A receiving module is used to receive input policy generation requests; wherein the policy generation request carries input data corresponding to the user; The recognition module is used to extract the input data from the policy generation request, and perform intent recognition on the input data based on a preset intent recognition agent to obtain the corresponding intent information; The retrieval module is used to perform knowledge base retrieval on the intent information based on a preset dual retrieval engine to obtain the corresponding retrieval results; The first processing module is used to perform tag recommendation processing on the search results based on a preset tag recommendation agent to obtain corresponding tag recommendation information; The second processing module is used to perform product generation processing on the intent information, the search results and the tag recommendation information based on a preset product generation intelligent agent to obtain the corresponding target product; The third processing module is used to perform strategy generation processing based on the intent information, the search results, the tag recommendation information and the target product to obtain the corresponding target strategy. The output module is used to process the output of the target strategy.

[0009] Thirdly, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described artificial intelligence-based strategy generation method.

[0010] Fourthly, a computer-readable storage medium is provided, which stores a computer program that, when executed by a processor, implements the steps of the above-described artificial intelligence-based strategy generation method.

[0011] In the above-mentioned scheme implemented by the AI-based strategy generation method, apparatus, computer equipment, and storage medium, a strategy generation request is first received; wherein the strategy generation request carries input data corresponding to the user; then, the input data is extracted from the strategy generation request, and the input data is subjected to intent recognition based on a preset intent recognition agent to obtain corresponding intent information; then, the intent information is subjected to knowledge base retrieval based on a preset dual retrieval engine to obtain corresponding retrieval results; subsequently, the retrieval results are subjected to tag recommendation processing based on a preset tag recommendation agent to obtain corresponding tag recommendation information; and the intent information, retrieval results, and tag recommendation information are subjected to product generation processing based on a preset product generation agent to obtain corresponding target products; further, the intent information, retrieval results, tag recommendation information, and target products are subjected to strategy generation processing to obtain corresponding target strategies; finally, the target strategies are output. Based on the above automated processing flow, unlike existing strategy generation methods, this application, through the combined use of an intent recognition agent, a dual retrieval engine, a tag recommendation agent, and a product generation agent, can intelligently and accurately complete the automated strategy generation processing of user input data, effectively improving the processing efficiency and accuracy of strategy generation, and ensuring the precision of the generated target strategy. Attached Figure Description

[0012] 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.

[0013] Figure 1 This is an exemplary system architecture diagram to which this application can be applied; Figure 2 This is a flowchart of an embodiment of the AI-based strategy generation method according to this application; Figure 3 This is a schematic diagram of the structure of an embodiment of the AI-based strategy generation apparatus according to this application; Figure 4 This is a schematic diagram of the structure of one embodiment of the computer device according to this application. Detailed Implementation

[0014] 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.

[0015] 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.

[0016] 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.

[0017] like Figure 1 As 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, etc.

[0018] 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.

[0019] 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.

[0020] 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.

[0021] It should be noted that the AI-based strategy generation method provided in this application is generally executed by a server / terminal device, and correspondingly, the AI-based strategy generation device is generally located in the server / terminal device.

[0022] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0023] Continue to refer to Figure 2 The flowchart illustrates an embodiment of the AI-based strategy generation method according to this application. The order of steps in the flowchart can be changed, and some steps can be omitted, depending on different needs. The AI-based strategy generation method provided in this application can be applied to any scenario requiring strategy generation, and therefore can be applied to products in these scenarios, such as strategy generation products in the financial insurance field. The AI-based strategy generation method includes the following steps: Step S201: Receive an input policy generation request; wherein the policy generation request carries input data corresponding to the user.

[0024] In this embodiment, the AI-based strategy generation method runs on an electronic device (e.g., Figure 1The server / terminal device shown can acquire policy generation requests input by the user via wired or wireless connection. It should be noted that the aforementioned wireless connection methods may include, but are not limited to, 3G / 4G / 5G connections, WiFi connections, Bluetooth connections, WiMAX connections, Zigbee connections, UWB (ultra-wideband) connections, and other currently known or future-developed wireless connection methods. The executing entity of this application is specifically a policy generation system, which can be simply referred to as the system. The system consists of an intent recognition agent (intent recognition agent module), a tag recommendation agent (tag recommendation agent module), and a product generation agent (product generation agent module). When an operator initiates a consultation or operation corresponding to a policy generation request through the system, the system will collect various input data in real time, i.e., the input data corresponding to the user. Input data includes, but is not limited to, the operator's click behavior on the interface, the input text content (such as search keywords, consultation questions), and the operation path (such as which page to jump to the consultation page). Here, the user can refer to the customer who needs consultation, or it can also refer to the operator themselves.

[0025] This application can be applied to strategy generation scenarios in the property insurance business within the fintech field. For example, staff can consult the system about car insurance-related issues corresponding to target customers, and the system will then perform corresponding strategy generation processing, including: User intent recognition: Through the intent recognition agent module, a hybrid expert model (MoE) is used to analyze user input information to determine whether the user's intent is to learn about car insurance product recommendations. Knowledge base retrieval: A dual-engine retrieval is adopted. Milvus performs semantic matching on unstructured user reviews and case analyses, while Elasticsearch retrieves structured car insurance product information, historical recommendation rules, etc., to obtain comprehensive relevant information. Tag recommendation: The tag recommendation agent module combines the user's historical data (such as the user having a previous small claim record and having consulted about a specific type of car insurance product) to recommend the tag "customer with a small claim record and interested in a specific product". Product generation: Based on the tags and knowledge base information, the product generation agent module generates a car insurance product combination for the user that includes a cost-effective option and specific value-added services. Strategy Generation: Based on the above information, the generated strategy is to push the recommended car insurance product combination information to the user via APP message after the user leaves work.

[0026] Step S202: Extract the input data from the policy generation request, and perform intent recognition on the input data based on a preset intent recognition agent to obtain the corresponding intent information.

[0027] In this embodiment, the required input data can be extracted from the policy generation request by parsing the aforementioned request. The specific implementation process of using a preset intent recognition agent to recognize the intent of the input data and obtain the corresponding intent information will be described in more detail in subsequent embodiments and will not be elaborated upon here.

[0028] Step S203: Based on the preset dual search engine, the intent information is searched in the knowledge base to obtain the corresponding search results.

[0029] In this embodiment, the specific implementation process of performing knowledge base retrieval on the intent information based on the preset dual retrieval engine to obtain the corresponding retrieval results will be further described in detail in subsequent specific embodiments of this application, and will not be elaborated on here.

[0030] Step S204: Based on the preset tag recommendation agent, perform tag recommendation processing on the search results to obtain the corresponding tag recommendation information.

[0031] In this embodiment, the specific implementation process of the above-mentioned tag recommendation agent performing tag recommendation processing on the search results to obtain the corresponding tag recommendation information will be further described in detail in subsequent specific embodiments of this application, and will not be elaborated on here.

[0032] Step S205: Based on the preset product generation intelligent agent, perform product generation processing on the intent information, the search results and the tag recommendation information to obtain the corresponding target product.

[0033] In this embodiment, the above-mentioned product generation agent performs product generation processing on the intent information, the search results and the tag recommendation information based on the preset product generation agent to obtain the corresponding target product. This application will describe this in more detail in subsequent specific embodiments, and will not elaborate further here.

[0034] Step S206: Based on the intent information, the search results, the tag recommendation information, and the target product, perform strategy generation processing to obtain the corresponding target strategy.

[0035] In this embodiment, the specific implementation process of generating a strategy based on the intent information, the search results, the tag recommendation information, and the target product to obtain the corresponding target strategy will be further described in detail in subsequent specific embodiments of this application, and will not be elaborated on here.

[0036] Step S207: Output processing is performed on the target strategy.

[0037] In this embodiment, the generated target strategy can be sent to the user via email, message, or other means, or presented to the user via a user interface, thereby completing the output processing of the target strategy.

[0038] This application first receives an input strategy generation request, wherein the strategy generation request carries input data corresponding to the user; then, the input data is extracted from the strategy generation request, and the input data is subjected to intent recognition based on a preset intent recognition agent to obtain corresponding intent information; then, the intent information is subjected to knowledge base retrieval based on a preset dual retrieval engine to obtain corresponding retrieval results; subsequently, the retrieval results are subjected to tag recommendation processing based on a preset tag recommendation agent to obtain corresponding tag recommendation information; and the intent information, retrieval results, and tag recommendation information are subjected to product generation processing based on a preset product generation agent to obtain corresponding target products; further, the intent information, retrieval results, tag recommendation information, and target products are subjected to strategy generation processing to obtain corresponding target strategies; finally, the target strategies are output. Based on the above automated processing flow, unlike existing strategy generation methods, this application, through the combined use of an intent recognition agent, a dual retrieval engine, a tag recommendation agent, and a product generation agent, can intelligently and accurately complete the automated strategy generation processing of user input data, effectively improving the processing efficiency and accuracy of strategy generation, and ensuring the precision of the generated target strategy.

[0039] In some optional implementations, step S202, which involves using a preset intent recognition agent to perform intent recognition on the input data to obtain corresponding intent information, includes the following steps: The intent recognition agent preprocesses the input data to obtain corresponding processed data.

[0040] In this embodiment, the preprocessing of input data includes: Data cleaning: Cleaning the input data to remove invalid, duplicate, or erroneous data. For example, if the text entered by the user contains a large number of irrelevant symbols or garbled characters, the system will filter and correct it to ensure the quality of data in subsequent processing. Data format conversion: Converting different types of data into a unified format for subsequent model processing. For example, converting user click behavior data into structured sequence data to facilitate model analysis.

[0041] The data is processed by extracting information based on a preset lightweight model to obtain the corresponding processing results.

[0042] In this embodiment, the information extraction process includes: Model selection: Selecting a lightweight model suitable for real-time data conversion and interaction scenarios. Lightweight models are characterized by fast computation speed and low resource consumption, enabling rapid response to user operations. Data input: Inputting the cleaned and format-converted processed data into the lightweight model. For example, for user-inputted text content, it is converted into a vector form that the model can process. Real-time processing: The lightweight model processes the input data in real time, quickly extracting key information. For example, for user search keywords, the model can identify the topic and intent of the keywords. Result output: Outputting the processed results (processing results) to provide preliminary information for subsequent intent recognition. For example, outputting the topic classification results of the user's search keywords.

[0043] The selection of the aforementioned lightweight models is not specifically limited; for example, lightweight text classification models (such as DistilBERT / TinyBERT), lightweight sequence labeling models (such as a simplified version of BiLSTM-CRF), lightweight semantic matching models (such as a simplified version of MobileBERT / Sentence-BERT), and so on, can be used. Intent analysis is performed on the processing results based on a pre-defined hybrid expert model to obtain multiple corresponding analysis results.

[0044] In this embodiment, the intent analysis includes: Expert model construction: Constructing multiple expert models (forming a hybrid expert model MoE), each expert model focusing on processing a specific type of user intent. For example, one expert model specializes in handling intents related to car insurance products, while another expert model specializes in handling intents related to claims consultation. Data allocation: Assigning the preliminary information output by the lightweight model to the corresponding expert model for processing. For example, if the lightweight model identifies that a user's search keywords are related to car insurance products, it assigns that information to the car insurance product expert model. Expert model processing: Each expert model, based on its own expertise and algorithms, performs in-depth analysis of the assigned data to identify more accurate user intents. For example, the car insurance product expert model analyzes the specific meaning and contextual information of the user's search keywords to determine whether the user wants to know about the price, coverage, or purchase process of car insurance products.

[0045] All the analysis results are then fused to obtain the corresponding fusion result.

[0046] In this embodiment, the processing results of each expert model are fused together, and the final user intent is derived by comprehensively considering the judgments of different expert models. For example, if both the auto insurance product expert model and the claims consultation expert model make certain judgments on the user's input, the system will combine the judgment results of the two models according to certain weights and rules to determine the user's main intent, i.e., the fused result.

[0047] The fusion result is used as the intent information.

[0048] In this embodiment, the identified user intent can be further verified against preset business rules to ensure the rationality and legality of the intent. For example, it can be checked whether the user intent conforms to the business scope and service policies of property insurance. If the identified user intent does not match the user's actual needs, the system can obtain more information through further interaction with the user and correct the intent. For example, the system can ask the user some questions to guide the user to clarify their needs, and then re-identify the intent based on the user's answers. In addition, the process and results of each intent recognition are recorded, analyzed, and summarized to continuously optimize the performance of the intent recognition model. For example, it can analyze which types of user intent are easily misidentified, identify the reasons, and make improvements.

[0049] Based on the above processing flow, this application preprocesses the input data using an intent-based intelligent agent to obtain corresponding processed data; then, it extracts information from the processed data using a pre-defined lightweight model to obtain corresponding processing results; subsequently, it performs intent analysis on the processing results using a pre-defined hybrid expert model to obtain multiple analysis results; finally, it fuses all analysis results to obtain a fusion result; and finally, it uses the fusion result as intent information. Thus, by combining a lightweight model and a hybrid expert model, this application can achieve efficient and accurate intent recognition of input data, ensuring the accuracy of the obtained intent information and providing an accurate intent foundation for subsequent precise strategy generation.

[0050] In some optional implementations of this embodiment, the dual search engine includes a first search engine and a second search engine; step S203 includes the following steps: The intent information is processed by intent transformation and query construction to obtain the corresponding vectorized query request and structured query statement.

[0051] In this embodiment, the intent conversion and query construction process includes: Intent conversion: converting the identified user intent (intent information) into a query statement suitable for knowledge base retrieval. For unstructured data retrieval, the user intent is converted into a semantic vector for vectorized matching in Milvus (the first search engine); for structured data retrieval, key information from the user intent is extracted and used as query keywords for Elasticsearch (the second search engine). Query construction: constructing a specific query statement based on the converted information. For Milvus, a vectorized query request is constructed, specifying the similarity threshold of the query vector and the number of returned results; for Elasticsearch, a structured query statement is constructed, such as using keyword matching, range queries, Boolean queries, etc.

[0052] The dual-engine initialization and configuration includes: Milvus engine (first search engine) initialization: Install and configure the Milvus vector database. Based on the scale of the unstructured data in the knowledge base and the retrieval requirements, set appropriate parameters, such as vector dimensions, storage space size, and index type. Simultaneously, perform performance tuning on Milvus to ensure efficient vectorized semantic matching. Elasticsearch engine (second search engine) initialization: Install and configure the Elasticsearch search engine. Create indexes, define their structure and fields to match the structured data in the knowledge base. Set appropriate shard and replica counts to improve search engine stability and query performance. Engine collaboration configuration: Establish a collaboration mechanism between Milvus and Elasticsearch to ensure they can cooperate for hybrid retrieval. For example, define data synchronization rules between different engines and how to integrate the search results from both engines during hybrid retrieval.

[0053] Based on the first search engine, the vectorized query request is retrieved in the preset knowledge base to obtain the corresponding first search result.

[0054] In this embodiment, the constructed vectorized query request is sent to the Milvus engine. The Milvus engine uses a vector similarity algorithm to search the unstructured data, finds the data most similar to the query vector, and sorts them according to similarity from high to low, thus obtaining the corresponding first search result.

[0055] The construction of the knowledge base includes: Data Collection: Collecting data related to property insurance business from multiple channels, including but not limited to historical strategy data, business rule documents, policy terms, user reviews, and case analyses. These data sources are wide-ranging, potentially originating from internal company records, market research reports, and customer feedback channels. Data Classification: Classifying the collected data into structured and unstructured data. Structured data, such as policy terms, has a clear format and fields; unstructured data, such as user reviews and case analyses, is more flexible and lacks a fixed format. Data Storage: Selecting appropriate storage methods based on data type. Structured data is stored in a relational database for easy structured querying and management; unstructured data is stored in a dedicated file system or object storage, with corresponding metadata established for easy subsequent retrieval and management. Data Updates: Regularly updating the data in the knowledge base to ensure its timeliness and accuracy. When new business rules are introduced, new cases are generated, or user feedback increases, this new data is promptly added to the knowledge base.

[0056] Based on the second retrieval engine, the structured query statement is retrieved in the knowledge base to obtain the corresponding second retrieval result.

[0057] In this embodiment, the structured query statement is sent to the Elasticsearch engine. The Elasticsearch engine performs a fast retrieval in the structured data based on the index and query conditions, returning results that meet the criteria, i.e., the second search result. During the dual-engine retrieval process, the retrieval progress and performance are monitored. If the retrieval speed of one engine is found to be too slow or abnormal, appropriate measures are taken promptly to adjust or repair it.

[0058] The first search result and the second search result are merged and sorted to obtain the corresponding target search result.

[0059] In this embodiment, the fusion and ranking process includes: fusing the search results from the two engines, comprehensively considering the semantic similarity of unstructured data and the keyword matching degree of structured data. For example, for the retrieval of car insurance product recommendation strategies, information related to car insurance product recommendations from user reviews and case analyses retrieved by Milvus is integrated with structured car insurance product information and historical recommendation rules retrieved by Elasticsearch. Then, the fused results are ranked according to certain ranking rules. The ranking rules can comprehensively consider factors such as the similarity, relevance, and importance of the results to ensure that the results returned to the user best match their intent.

[0060] The target search result is used as the search result.

[0061] Based on the above processing flow, this application transforms intent information and constructs queries to obtain corresponding vectorized query requests and structured query statements. Then, based on a first search engine, it searches the preset knowledge base for the vectorized query requests to obtain corresponding first search results. Similarly, based on a second search engine, it searches the knowledge base for the structured query statements to obtain corresponding second search results. The first and second search results are then fused and sorted to obtain the corresponding target search results. These target search results are subsequently used as the final search results. Thus, by combining the first and second search engines, this application can comprehensively and accurately complete the knowledge base retrieval processing of intent information, improving the accuracy and comprehensiveness of the generated search results.

[0062] In some optional implementations, the label recommendation agent includes multiple label recommendation sub-agents; step S204 includes the following steps: Obtain the user's historical data.

[0063] In this embodiment, historical user data can be collected from the insurance company's user database. This data covers multiple aspects such as the user's claims history, coverage preferences, purchase records, and consultation records. For example, the user's claims history records the number of claims, the amount of claims, and the reasons for claims; coverage preferences can be determined by the coverage amounts the user has purchased in the past and the coverage range mentioned during consultations.

[0064] The historical data and the search results are integrated and processed to obtain the corresponding target data.

[0065] In this embodiment, search results retrieved from the knowledge base and user historical data are integrated to form a complete dataset (i.e., target data), providing comprehensive data support for subsequent tag recommendations. For example, car insurance product information can be combined with users' coverage preferences to better recommend suitable tags to users.

[0066] The target data is processed by multiple tag recommendation sub-agents to obtain multiple corresponding tag information.

[0067] In this embodiment, the Multi-Agent collaborative architecture construction includes: **Label Recommendation Agent Module Construction (i.e., Label Recommendation Intelligent Agent):** Determining the composition and functions of the Label Recommendation Agent module. This module consists of multiple sub-Agents, each responsible for different label recommendation tasks, such as a claims record analysis agent, a coverage preference analysis agent, and a product interest analysis agent. **Sub-Agent Function Definition:** Clarifying the specific functions and workflows of each sub-Agent. For example, the claims record analysis agent is responsible for analyzing a user's claims history to determine their risk level; the coverage preference analysis agent determines the user's coverage preference type based on their coverage selection and consultation. **Agent Communication Mechanism Establishment:** Establishing a communication mechanism between sub-Agents to ensure they can collaborate and share information. For example, the claims record analysis agent can transmit analysis results to the coverage preference analysis agent, allowing the latter to comprehensively consider the user's risk and coverage preference when making label recommendations.

[0068] Specifically, the sub-Agent tag recommendation process includes: Claims Record Analysis Agent: Data Extraction: Extracts users' historical claims data from the integrated dataset, including the number of claims, claim amount, and reason for claim. Risk Assessment: Assesses the user's risk based on the claims data. For example, users with more claims and larger claim amounts have a higher risk level. Tag Recommendation: Recommends corresponding risk tags to users based on the risk assessment results, such as "high-risk customer" or "low-risk customer". Coverage Preference Analysis Agent: Data Extraction: Extracts users' coverage selection data and consultation data to understand users' coverage preferences. Preference Classification: Classifies users' coverage preferences, such as "high coverage preference", "low coverage preference", or "medium coverage preference". Tag Recommendation: Recommends corresponding coverage preference tags to users based on the preference classification results. Product Interest Analysis Agent: Data Extraction: Analyzes users' consultation and purchase records to understand users' interests in different types of auto insurance products. Interest Identification: Identifies the degree of user interest in specific auto insurance products, such as "strong interest", "moderate interest", or "no interest". Tag Recommendation: Based on the interest identification results, relevant product interest tags are recommended to users, such as "vehicle damage insurance interested customers" and "third-party liability insurance interested customers".

[0069] All the aforementioned tag information is subjected to tag fusion and optimization processing to obtain the corresponding specified tag information.

[0070] In this embodiment, the tag fusion and optimization process includes: Tag fusion: fusing the tags recommended by each sub-Agent, comprehensively considering the importance and relevance of different tags. For example, for users who simultaneously have the tags "high-risk customer" and "high coverage preference," further analysis is conducted on the correlation between these two tags to determine whether the tag weights need to be adjusted or new tags added. Tag optimization: optimizing the fused tags according to business rules and actual needs. For example, if certain tag combinations are uncommon or unreasonable in actual business, they can be adjusted or merged. Tag verification: verifying the optimized tags against the user's actual situation to ensure the accuracy and effectiveness of the tags. For example, by further communication with the user or observing the user's subsequent behavior, verifying whether the tags can accurately reflect the user's characteristics and needs.

[0071] The specified tag information is used as tag recommendation information.

[0072] Based on the above processing flow, this application obtains the user's historical data; integrates the historical data with the search results to obtain corresponding target data; then, based on multiple tag recommendation sub-agents, performs tag recommendation processing on the target data to obtain multiple corresponding tag information; subsequently, all the tag information undergoes tag fusion and optimization processing to obtain corresponding specified tag information; finally, the specified tag information is used as tag recommendation information. Thus, by using a tag recommendation agent composed of multiple tag recommendation sub-agents, this application can automatically and accurately complete the tag recommendation processing of target data obtained by integrating historical data and search results, improving the accuracy of the generated tag recommendation information.

[0073] In some alternative implementations, step S205 includes the following steps: Collect historical successful case studies and market data.

[0074] In this embodiment, successful product recommendation cases can be compiled and analyzed. This allows us to understand which product combinations or value-added services are popular among different user groups and the reasons for their success, thus obtaining corresponding historical success stories. For example, certain combinations of auto insurance products may sell well among young working professionals because their prices are reasonable and their coverage meets the needs of this group. Furthermore, it's important to monitor the dynamics of the current insurance market (such as the auto insurance market) and the value-added services market. Understanding competitors' product features, pricing strategies, and market trends provides market data. For instance, recent launches of new auto insurance add-ons, such as exclusive protection for new energy vehicles, can help provide more market-demand-oriented options when generating new products.

[0075] Based on the search results, the historical successful recommendation cases, and the market situation data, the product generation agent is used to perform product matching processing on the intent information and the tag recommendation information to obtain the corresponding initial products.

[0076] In this embodiment, the aforementioned product generation intelligent agent (product generation Agent module) consists of multiple sub-Agents (product generation sub-agents), such as product matching agents, market trend analysis agents, and user demand analysis agents. The sub-Agent functional definitions include: clearly defining the specific functions and workflows of each sub-Agent. For example, the product matching agent is responsible for filtering potentially suitable products for users based on user tags and product information from the knowledge base; the market trend analysis agent analyzes the current market situation to provide a reference for product combinations; and the user demand analysis agent combines users' historical data and consultation records to gain a deeper understanding of user needs. An inter-Agent communication mechanism is established: a communication mechanism is established between sub-Agents to ensure they can collaborate and share information. For example, the product matching agent can pass the initially filtered product information to the market trend analysis agent, which adjusts and optimizes the products based on market conditions, and then passes the results to the user demand analysis agent for further analysis.

[0077] In addition, the product generation process based on sub-Agents (product generation sub-agents) includes: For the product matching agent, tag matching: Based on the user's target customer group tags, products matching the tags are filtered from the product information in the knowledge base. For example, for users tagged "high-risk, high-coverage customers," auto insurance products with broad coverage and high coverage are filtered. Preliminary screening: The filtered products are preliminarily screened to exclude products that do not meet the basic conditions. For example, if the user's vehicle is old, some products with age restrictions may be excluded. Product list generation: A preliminary product list is generated, containing basic product information and characteristics, for further analysis by the agent. For the market trend analysis agent, market data collection: Data on the current auto insurance market and value-added services market is collected, including product prices, sales volume, user reviews, etc. Trend analysis: Market trends are analyzed to understand which products or services are competitive in the current market. For example, if the new energy vehicle market is growing rapidly, insurance products and services related to new energy vehicles may be more popular. Product adjustment: The product list generated by the product matching agent is adjusted based on the market trend analysis results. For example, adding new products that align with market trends or adjusting the combination of existing products. For user needs analysis agents, user data integration involves integrating historical user data and consultation records to gain a deeper understanding of user needs and preferences. For example, does the user frequently inquire about specific types of value-added services, or how price-sensitive are they? Needs matching involves matching the adjusted product list with user needs to determine which products best suit the user's actual requirements. For example, if a user frequently inquires about free roadside assistance, then car insurance products that include this service should be prioritized in the product portfolio. Personalized recommendations involve generating personalized product recommendation plans for the user based on the user needs matching results. For example, recommending a car insurance product portfolio that includes multiple types of insurance, high coverage, and free roadside assistance for a "high-risk, high-coverage customer."

[0078] The initial goods are optimized based on a preset optimization strategy to obtain the optimized specified goods.

[0079] In this embodiment, the optimization of the initial goods includes: 1. Combination rationality assessment: The rationality of the generated goods combination is assessed, considering the complementarity and coordination between products. For example, whether there are conflicts or overlaps between car insurance products and value-added services, and whether they can provide comprehensive protection for users. Targeted optimization is performed when unreasonableness is detected, including: 1) Conflict-related optimization: Unified rules and clear boundary operations: Clause alignment: Ensure that the clause logic of all products within the combination is consistent (e.g., exclusions, waiting periods, claims conditions). Liability segmentation: Clarify the core coverage of each product to avoid overlapping coverage. Example: In a car insurance combination, if the main insurance excludes "non-designated driver," then the value-added service "designated driver insurance" must be limited to "only covering accidents during the designated driver service period," and clearly marked "not subject to duplicate claims with the main insurance." In a health insurance combination, the waiting period for all products is unified to 30 days, or the waiting period for specific coverage is extended through supplementary insurance. 2) Redundancy Optimization: Merging Redundancy and Layered Design: Functional Integration: Merging duplicate coverage or services into a single product, tiered by coverage level (e.g., basic version, upgraded version). Price Differentiation: Differentiating product positioning by adjusting coverage amount, service frequency, or scope. Example: In a car insurance package, merging "Basic Roadside Assistance" (3 times / year) and "High-End Assistance" (unlimited times + hotel subsidy) into a "Assistance Service Package," allowing users to choose according to their needs. In a travel insurance package, integrating "Medical Expense Reimbursement" and "Emergency Medical Transport" into a "Medical Protection Plan," offering different coverage options (e.g., 100,000 / 300,000). 3) Gap-Filling Optimization: Supplementing Core Needs and Improving Coverage: Needs Research: Analyzing high-frequency, under-covered scenarios through user feedback or claims data (e.g., fractures in the elderly, battery damage in new energy vehicles). Product Expansion: Adding targeted products or supplementary insurance to the package. Example: Adding "New Energy Vehicle Battery Insurance" to a car insurance package, covering risks such as battery spontaneous combustion and collision damage. Adding "Fracture-Specific Coverage" to an elderly accident insurance package, providing high compensation or rehabilitation services. 2. Price Optimization: Optimize the pricing of the product mix based on market prices and user needs. For example, if users are price-sensitive, consider adjusting the coverage amount or service content of the product to reduce the total price. 3. Risk Assessment: Assess the potential risks to the product mix, such as claims risk and market competition risk. Based on the risk assessment results, further adjust and optimize the product mix.

[0080] The specified goods are designated as the target goods.

[0081] Based on the above processing flow, this application collects historical successful recommendation cases and market data; then, based on the search results, historical successful recommendation cases, and market data, it uses a product generation intelligent agent to match intent information and tag information to obtain corresponding initial products; subsequently, it optimizes the initial products based on a preset optimization strategy to obtain optimized designated products; finally, it uses the designated products as target products. Thus, by using a product generation intelligent agent to match intent information and tag information based on search results, historical successful recommendation cases, and market data, this application can intelligently and accurately generate products that meet user needs and are suitable for recommendation, ensuring the accuracy of the generated target products.

[0082] In some optional implementations of this embodiment, step S206 includes the following steps: The intent information, the search results, the tag recommendation information, and the target product are integrated to obtain corresponding integrated information.

[0083] In this embodiment, the above information integration includes: User Intent Analysis: Reviewing and analyzing the user's initial expressed needs and intentions to clarify the services or problems the user expects to obtain through property insurance. For example, the user may want to purchase a cost-effective and comprehensive car insurance policy for their vehicle, or they may want to learn about specific types of value-added services. Knowledge Base Information Summary: Summarizing and organizing various types of information (search results) previously retrieved from the knowledge base, such as car insurance product characteristics, historical strategy cases, business rules, etc., to ensure that this information is presented clearly and comprehensively for reference in subsequent strategy formulation. Tag Information Integration: Integrating and analyzing the tags generated for users by the tag recommendation agent module. Understanding the user characteristics and needs represented by each tag, for example, the tag "high-risk, high-coverage customer" means that the user has specific attributes in terms of risk assessment and coverage preferences. Product Information Integration: Organizing in detail the product information generated for users by the product generation agent module, including recommended car insurance product combinations, value-added service items, etc., to clarify the characteristics, advantages, and applicable scenarios of each product. Then, all the intent information, search results, tag recommendation information and target products after integration or sorting are integrated to obtain the corresponding integrated information.

[0084] Obtain the preset strategy framework.

[0085] In this embodiment, the construction of the strategy framework includes: Determining strategy objectives: Based on user intent and business goals, determine the main objectives of the strategy. For example, for users who want to purchase car insurance, the strategy objective might be to increase the conversion rate of users purchasing recommended car insurance products; for users inquiring about value-added services, the objective might be to increase users' understanding of value-added services and their willingness to purchase. Dividing strategy dimensions: Constructing the strategy framework from multiple dimensions, such as product dimension, channel dimension, and time dimension. In the product dimension, consider the combination of recommended car insurance products and value-added services; in the channel dimension, determine which channels to use to recommend the strategy to users, such as Ping An Property & Casualty Insurance APP, SMS, telephone, etc.; in the time dimension, plan the time nodes and frequency of recommendations. Formulating strategy rules: Based on business rules and logic, formulate specific rules for the strategy. For example, for "high-risk, high-coverage customers," specify which types of insurance must be included in the recommended car insurance product combination, and the minimum coverage requirement; in terms of channel selection, prioritize recommendations via APP push notifications and SMS.

[0086] Based on the aforementioned strategy framework, the integrated information is processed to generate a corresponding initial strategy.

[0087] In this embodiment, the strategy generation process includes: Product recommendation content: Based on the product generation results, a detailed description of the recommended auto insurance product combinations and value-added services is provided. This includes information such as the coverage, price, advantages, and features of each product, as well as how the product combination meets the user's needs. For example, for "high-risk, high-coverage customers," the recommended auto insurance product combination may include vehicle damage insurance, third-party liability insurance, theft insurance, etc., with high coverage and free roadside assistance. Recommendation channel selection: Based on the user's behavior and preferences, appropriate recommendation channels are selected. For example, if the user frequently uses the Ping An Property & Casualty Insurance APP, recommendations are prioritized through APP push notifications and in-app pop-ups; if the user has a high acceptance of SMS messages, SMS reminders can also be used. Recommendation time planning: The timing and frequency of recommendations are determined. For example, recommendation information is pushed immediately when the user logs into the APP, or recommendations are concentrated during specific time periods (such as before the user's vehicle annual inspection). At the same time, the frequency of recommendations is controlled to avoid harassing the user. Supporting service description: In addition to the product recommendations themselves, related supporting services should also be provided to the user. For example, providing users with purchase guides for car insurance products, explanations of the claims process, or instructions on how to use value-added services can enhance user experience and trust.

[0088] The initial strategy is evaluated and optimized to obtain the optimized specified strategy.

[0089] In this embodiment, the above evaluation and optimization process includes: Simulation evaluation: Before formally implementing the strategy, the strategy is evaluated through simulation. The simulation simulates the reactions of different user groups to the strategy, predicting the potential effects, such as purchase conversion rate and user satisfaction. Based on the simulation evaluation results, the strategy is adjusted and optimized. Small-scale testing: A representative group of users is selected for small-scale testing to observe the actual user reactions to the strategy. User feedback is collected to understand user satisfaction with the recommended products, recommendation channels, and recommendation timing, as well as whether there are other problems and needs. Strategy optimization: Based on the results of the small-scale testing, the strategy is further optimized. For example, if it is found that users have a low acceptance of a certain recommendation channel, the channel strategy can be adjusted; if users are highly price-sensitive, the product mix can be appropriately adjusted or promotional activities can be offered.

[0090] The specified strategy is used as the target strategy.

[0091] In this embodiment, the optimized strategy can be formally implemented. Strategy information is pushed to the target user group according to predetermined recommendation channels, times, and content. The accuracy and timeliness of the implementation process are ensured to avoid errors or delays. During strategy implementation, relevant data, such as the number of times the recommendation information is displayed, click-through rate, and purchase conversion rate, are monitored in real time. Data analysis is used to understand the effectiveness of the strategy and identify potential problems in a timely manner. Furthermore, the strategy is dynamically adjusted based on the data monitoring results. If the effect of a certain recommendation channel is found to be poor, the channel strategy can be adjusted promptly; if users have a high interest in a certain product, the recommendation frequency of that product can be appropriately increased. Simultaneously, continuous attention is paid to market changes and changes in user needs, and the strategy content is updated in a timely manner to ensure that the strategy remains targeted and effective.

[0092] Based on the above processing flow, this application integrates intent information, search results, tag recommendation information, and target products to obtain corresponding integrated information; then, it obtains a preset strategy framework; and based on the strategy framework, it performs strategy generation processing on the integrated information to obtain a corresponding initial strategy; subsequently, it evaluates and optimizes the initial strategy to obtain an optimized specified strategy; finally, it uses the specified strategy as the target strategy. This application can intelligently and accurately construct a complete target strategy by integrating intent information, search results, tag recommendation information, and target products, ensuring that the target strategy accurately meets user needs, achieves business objectives, and improves the accuracy and adaptability of the generated target strategy.

[0093] In some optional implementations of this embodiment, after step S206, the electronic device may further perform the following steps: Collect end-to-end data of the target strategy during its execution.

[0094] In this embodiment, end-to-end data during the strategy execution process can be collected based on an A / B testing platform and a big data platform. This includes push data (push time, push content, push channel, etc.), reach data (time of reaching user devices, device type, etc.), click data (click time, clicked page, etc.), and completion data (time of purchase or transaction completion, amount, etc.). The collected data is then integrated to establish a full-link data tracking system for strategy execution. For example, push data is correlated with reach data to determine which push messages successfully reached users; then reach data is correlated with click data to determine which reached users clicked on the strategy information; finally, click data is correlated with completion data to determine which clicked users completed the strategy's target behavior.

[0095] Anomaly detection is performed on the entire data based on a preset threshold to obtain the corresponding anomaly detection results.

[0096] In this embodiment, reasonable thresholds are set for each metric based on historical data and business objectives. For example, the normal range for push notification reach rate is set to 70%-90%, the normal range for click-through rate is 8%-25%, and the normal range for completion rate is 3%-15%. Anomaly detection includes real-time monitoring of each metric's value; when a metric value exceeds the set threshold range, it is determined to be an anomaly. For example, if the completion rate of a car insurance recommendation strategy suddenly drops to 1%, below the lower limit of the normal range, it is determined to be an abnormal completion rate.

[0097] Root cause analysis is performed on the entire data based on a preset intelligent attribution model to obtain the corresponding root cause analysis results.

[0098] In this embodiment, a pre-built intelligent attribution model can be used to analyze the entire data chain, locate the root cause of anomalies, and output the corresponding root cause analysis results. For example, if the completion rate is abnormal, the analysis will determine whether the decrease in click-through rate is due to a sudden drop in reach rate or a decrease in click-through rate; if the reach rate drops sharply, further analysis will determine whether the problem lies with the push channel or with abnormal reception on the user's device.

[0099] The anomaly detection results and the root cause analysis results are sent to relevant personnel.

[0100] In this embodiment, the generated anomaly detection results and root cause analysis results can be sent to relevant personnel via email or text message. These relevant personnel can be business personnel involved in the operation and maintenance work related to policy execution.

[0101] In addition, the system can also combine the configuration ideas of high-scoring, excellent strategies regarding people, goods, and places, compare and summarize the current strategy configuration, and output optimization suggestions. Specifically, it compares the current strategy configuration (people, goods, and places) with high-scoring, excellent strategies. For example, it compares the configuration of recommended car insurance product combinations, recommendation channels, and recommendation times. Then, it summarizes the differences between the current strategy and excellent strategies, identifying the problems and shortcomings of the current strategy. For example, it finds that the current strategy's car insurance product combination is not rich enough and lacks attractive value-added services. Finally, based on the difference analysis results, it outputs specific optimization suggestions. For example, it suggests adjusting the car insurance product combination, adding some popular insurance types or value-added services; or it suggests optimizing recommendation channels, increasing the proportion of SMS or telephone recommendations, etc.

[0102] Based on the above processing flow, this application collects end-to-end data of the target strategy during its execution; then, it performs anomaly detection on the end-to-end data based on preset thresholds to obtain corresponding anomaly detection results; and finally, it performs root cause analysis on the end-to-end data based on a preset intelligent attribution model to obtain corresponding root cause analysis results. Subsequently, the anomaly detection results and the root cause analysis results are sent to relevant personnel. Thus, after collecting end-to-end data of the target strategy during its execution, this application can achieve intelligent anomaly detection of the end-to-end data based on preset thresholds, and intelligent root cause analysis of the end-to-end data based on an intelligent attribution model, improving the processing efficiency and intelligence of anomaly detection and root cause analysis of the strategy.

[0103] In some optional implementations, the system also features a comprehensive scoring system based on push reach rate, click-through rate, and completion rate, using piecewise linear interpolation to quantify task quality and address the issue of bias in single-indicator evaluation. Specifically, this includes: 1. Multi-dimensional indicator fusion. Indicator determination: Push reach rate: The percentage of strategy information successfully delivered to users through different recommendation channels (such as Ping An Insurance APP push notifications, SMS, etc.). For example, for APP push notifications, the total number of push messages and the number of successfully delivered user devices are recorded to calculate the reach rate. Click-through rate: Measures the user's interest in the recommended strategy information, calculating the ratio of the number of times users click to view strategy details or related pages to the number of users reached. For example, after an APP push notification, the number of users who clicked the message to enter the car insurance product recommendation page is recorded. Completion rate: Reflects the percentage of users who ultimately complete the strategy's target behavior, such as purchasing recommended car insurance products or handling value-added services. The ratio of the number of users who completed the purchase or handling to the number of users who clicked to view strategy details is calculated. Piecewise linear interpolation quantification. Setting segment intervals: Based on historical data and business experience, different segment intervals are set for each indicator. For example, for push notification reach rate, four intervals are set: below 60%, 60%-80%, 80%-95%, and above 95%; for click-through rate (CTR), four intervals are set: below 5%, 5%-15%, 15%-30%, and above 30%; for completion rate, four intervals are set: below 2%, 2%-10%, 10%-25%, and above 25%. Score allocation: A corresponding score is assigned to each interval, with higher scores indicating better performance for that metric. For example, a push notification reach rate below 60% receives 1 point, 60%-80% receives 3 points, 80%-95% receives 5 points, and above 95% receives 7 points; CTR and completion rate are assigned scores in a similar manner. Calculating the overall score: Based on the scores of each metric and the preset weights, the overall score of the strategy is calculated. For example, assuming the push reach rate has a weight of 0.3, the click-through rate has a weight of 0.4, and the completion rate has a weight of 0.3, the overall score = push reach rate score × 0.3 + click-through rate score × 0.4 + completion rate score × 0.3.

[0104] 2. Contextualized Weight Adaptation. Channel and Task Type Identification: Channel Identification: Identify the recommendation channel used for the push notification, such as APP messages, SMS, phone calls, etc. Different channels have different characteristics and user acceptance levels, therefore, the indicator weights need to be adjusted according to the channel characteristics. Task Type Identification: Determine the task type of the strategy, such as car insurance recommendations, value-added service processing, customer maintenance tasks, etc. Different task types have different goals and focuses, requiring corresponding adjustments to the indicator weights. Weight Configuration: Car Insurance Recommendation Strategy via APP Message Push: Since APP message pushes require users to actively click and view, click-through rate and completion rate better reflect the effectiveness of the strategy, therefore, the weights of these two indicators should be appropriately increased. For example, the push reach rate weight is set to 0.2, the click-through rate weight to 0.5, and the completion rate weight to 0.3. Customer Maintenance Task Related Strategies: Customer maintenance tasks focus more on continuous interaction with users and user engagement, making reach rate and user engagement more important. Therefore, the weights of push reach rate and user engagement indicators should be appropriately increased. For example, push notification reach rate is weighted at 0.4, user engagement metrics (such as the percentage of users replying to SMS messages or answering phone calls) are weighted at 0.4, and other metrics are weighted at 0.2.

[0105] In some alternative implementations, the user information obtained is subject to user consent and complies with relevant laws and policies.

[0106] Furthermore, any software tools or components not belonging to our company that appear in the embodiments of this application are merely illustrative examples and do not represent actual use.

[0107] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0108] It should be emphasized that, to further ensure the privacy and security of the aforementioned target strategy, the target strategy can also be stored in a node of a blockchain.

[0109] 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.

[0110] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use 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. 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.

[0111] 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 above methods. 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).

[0112] 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.

[0113] Further reference Figure 3 As a response to the above Figure 2 To implement the method shown, this application provides an embodiment of an artificial intelligence-based strategy generation device, which is similar to... Figure 2 Corresponding to the method embodiments shown, this device can be specifically applied to various electronic devices.

[0114] like Figure 3As shown, the AI-based strategy generation device 300 described in this embodiment includes: a receiving module 301, an identification module 302, a retrieval module 303, a first processing module 304, a second processing module 305, a third processing module 306, and an output module 307. Wherein: The receiving module 301 is used to receive an input policy generation request; wherein the policy generation request carries input data corresponding to the user; The identification module 302 is used to extract the input data from the policy generation request, and perform intent identification on the input data based on a preset intent identification agent to obtain the corresponding intent information; The retrieval module 303 is used to perform a knowledge base retrieval on the intent information based on a preset dual retrieval engine to obtain the corresponding retrieval results; The first processing module 304 is used to perform tag recommendation processing on the search results based on a preset tag recommendation agent to obtain corresponding tag recommendation information; The second processing module 305 is used to perform product generation processing on the intent information, the search results and the tag recommendation information based on a preset product generation intelligent agent to obtain the corresponding target product; The third processing module 306 is used to perform strategy generation processing based on the intent information, the search results, the tag recommendation information and the target product to obtain the corresponding target strategy. The output module 307 is used to perform output processing on the target strategy.

[0115] In some optional implementations of this embodiment, the identification module 302 includes: The preprocessing submodule is used to preprocess the input data based on the intent recognition agent to obtain the corresponding processed data; The extraction submodule is used to extract information from the processed data based on a preset lightweight model to obtain the corresponding processing results. The analysis submodule is used to perform intent analysis on the processing results based on a preset hybrid expert model to obtain multiple corresponding analysis results; The fusion submodule is used to fuse all the analysis results to obtain the corresponding fusion result; The first determining submodule is used to use the fusion result as the intent information.

[0116] In some optional implementations of this embodiment, the dual search engine includes a first search engine and a second search engine; the search module 303 includes: A submodule is constructed to perform intent transformation and query construction processing on the intent information to obtain the corresponding vectorized query request and structured query statement; The first retrieval submodule is used to retrieve the vectorized query request from a preset knowledge base based on the first retrieval engine to obtain the corresponding first retrieval result; The second retrieval submodule is used to retrieve the structured query statement from the knowledge base based on the second retrieval engine to obtain the corresponding second retrieval result; The first processing submodule is used to fuse and sort the first search result and the second search result to obtain the corresponding target search result. The second determining submodule is used to use the target retrieval result as the retrieval result.

[0117] In some optional implementations of this embodiment, the tag recommendation agent includes multiple tag recommendation sub-agents; the first processing module 304 includes: The first acquisition submodule is used to acquire the user's historical data; The first integration submodule is used to integrate the historical data and the search results to obtain the corresponding target data; The recommendation submodule is used to perform label recommendation processing on the target data based on multiple label recommendation sub-agents to obtain corresponding multiple label information; The second processing submodule is used to perform tag fusion and optimization processing on all the tag information to obtain the corresponding specified tag information; The third determining submodule is used to use the specified tag information as tag recommendation information.

[0118] In some optional implementations of this embodiment, the second processing module 305 includes: The collection submodule is used to collect historical successful recommendation cases and market situation data; The matching submodule is used to perform product matching processing on the intent information and the tag recommendation information based on the search results, the historical successful recommendation cases and the market situation data, using the product generation agent to obtain the corresponding initial products. The optimization submodule is used to optimize the initial goods based on a preset optimization strategy to obtain the optimized specified goods. The fourth determination submodule is used to select the specified goods as the target goods.

[0119] In some optional implementations of this embodiment, the third processing module 306 includes: The second integration submodule is used to integrate the intent information, the search results, the tag recommendation information and the target product to obtain corresponding integrated information. The second acquisition submodule is used to acquire the preset strategy framework; The third processing submodule is used to perform policy generation processing on the integrated information based on the policy framework to obtain the corresponding initial policy. The fourth processing submodule is used to evaluate and optimize the initial strategy to obtain the optimized specified strategy; The fifth determining submodule is used to use the specified strategy as the target strategy.

[0120] In some optional implementations of this embodiment, the AI-based strategy generation device further includes: The collection module is used to collect end-to-end data of the target policy during the policy execution process; The detection module is used to perform anomaly detection on the entire link data based on a preset threshold and obtain the corresponding anomaly detection results; The analysis module is used to perform root cause analysis on the entire data based on a preset intelligent attribution model, and obtain the corresponding root cause analysis results. The sending module is used to send the anomaly detection results and the root cause analysis results to relevant personnel.

[0121] In this embodiment, the operations performed by the above modules or units correspond one-to-one with the steps of the artificial intelligence-based strategy generation method in the aforementioned implementation method, and will not be repeated here. 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.

[0122] 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 components 41-43 is shown in the figure; however, it should be understood that it is not required to implement all the shown components, 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.

[0123] 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.

[0124] 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 strategy generation methods based on artificial intelligence. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or will be output.

[0125] In some embodiments, the processor 42 may be a central processing unit (CPU), a controller, a microcontroller, a 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, for example, to execute computer-readable instructions of the artificial intelligence-based strategy generation method.

[0126] 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.

[0127] This application also provides another embodiment, namely, providing 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 artificial intelligence-based strategy generation method described above.

[0128] 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.

[0129] 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 strategy generation method based on artificial intelligence, characterized in that, Includes the following steps: Receive a policy generation request; wherein the policy generation request carries input data corresponding to the user; The input data is extracted from the policy generation request, and the input data is subjected to intent recognition based on a preset intent recognition agent to obtain the corresponding intent information; The intent information is retrieved from the knowledge base based on a preset dual search engine to obtain the corresponding search results; The search results are processed by a pre-defined tag recommendation agent to obtain corresponding tag recommendation information. Based on a pre-defined intelligent agent for generating goods, the agent processes the intent information, the search results, and the tag recommendation information to generate the corresponding target goods. Based on the intent information, the search results, the tag recommendation information, and the target product, a strategy generation process is performed to obtain the corresponding target strategy. The target strategy is then processed for output.

2. The strategy generation method based on artificial intelligence according to claim 1, characterized in that, The step of performing intent recognition on the input data based on a preset intent recognition agent to obtain corresponding intent information specifically includes: The intent recognition agent preprocesses the input data to obtain corresponding processed data. The data is processed by extracting information based on a preset lightweight model to obtain the corresponding processing results. The processing results are subjected to intent analysis based on a preset hybrid expert model to obtain multiple corresponding analysis results; All the analysis results are fused to obtain the corresponding fusion result; The fusion result is used as the intent information.

3. The strategy generation method based on artificial intelligence according to claim 1, characterized in that, The dual search engine includes a first search engine and a second search engine; the step of performing a knowledge base search on the intent information based on the preset dual search engine to obtain the corresponding search results specifically includes: The intent information is processed by intent transformation and query construction to obtain the corresponding vectorized query request and structured query statement; Based on the first search engine, the vectorized query request is retrieved in the preset knowledge base to obtain the corresponding first search result; Based on the second retrieval engine, the structured query statement is retrieved in the knowledge base to obtain the corresponding second retrieval result; The first search result and the second search result are merged and sorted to obtain the corresponding target search result; The target search result is used as the search result.

4. The strategy generation method based on artificial intelligence according to claim 1, characterized in that, The tag recommendation agent includes multiple tag recommendation sub-agents; the step of performing tag recommendation processing on the search results based on the preset tag recommendation agent to obtain corresponding tag recommendation information specifically includes: Obtain the user's historical data; The historical data and the search results are integrated and processed to obtain the corresponding target data; Based on multiple label recommendation sub-agents, the target data is processed to perform label recommendation, thereby obtaining multiple corresponding label information; All the aforementioned tag information is subjected to tag fusion and optimization processing to obtain the corresponding specified tag information; The specified tag information is used as tag recommendation information.

5. The strategy generation method based on artificial intelligence according to claim 1, characterized in that, The step of the intelligent agent for generating goods based on a preset information to generate goods from the intent information, the search results, and the tag recommendation information to obtain the corresponding target goods specifically includes: Collect historical successful recommendation cases and market data; Based on the search results, the historical successful recommendation cases, and the market situation data, the product generation agent is used to perform product matching processing on the intent information and the tag recommendation information to obtain the corresponding initial products. The initial goods are optimized based on a preset optimization strategy to obtain the optimized specified goods. The specified goods are designated as the target goods.

6. The strategy generation method based on artificial intelligence according to claim 1, characterized in that, The step of generating a corresponding target strategy based on the intent information, the search results, the tag recommendation information, and the target product specifically includes: The intent information, the search results, the tag recommendation information, and the target product are integrated to obtain corresponding integrated information. Obtain the preset strategy framework; Based on the aforementioned strategy framework, the integrated information is processed to generate a corresponding initial strategy. The initial strategy is evaluated and optimized to obtain the optimized specified strategy; The specified strategy is used as the target strategy.

7. The strategy generation method based on artificial intelligence according to claim 1, characterized in that, After the step of outputting the target policy, the method further includes: Collect end-to-end data of the target strategy during its execution process; Anomaly detection is performed on the entire link data based on a preset threshold to obtain the corresponding anomaly detection results; Root cause analysis is performed on the entire data based on a preset intelligent attribution model to obtain the corresponding root cause analysis results. The anomaly detection results and the root cause analysis results are sent to relevant personnel.

8. A strategy generation device based on artificial intelligence, characterized in that, include: A receiving module is used to receive input policy generation requests; wherein the policy generation request carries input data corresponding to the user; The recognition module is used to extract the input data from the policy generation request, and perform intent recognition on the input data based on a preset intent recognition agent to obtain the corresponding intent information; The retrieval module is used to perform knowledge base retrieval on the intent information based on a preset dual retrieval engine to obtain the corresponding retrieval results; The first processing module is used to perform tag recommendation processing on the search results based on a preset tag recommendation agent to obtain corresponding tag recommendation information; The second processing module is used to perform product generation processing on the intent information, the search results and the tag recommendation information based on a preset product generation intelligent agent to obtain the corresponding target product; The third processing module is used to perform strategy generation processing based on the intent information, the search results, the tag recommendation information and the target product to obtain the corresponding target strategy. The output module is used to process the output of the target strategy.

9. A computer device, characterized in that, The system includes a memory and a processor, wherein the memory stores computer-readable instructions, and the processor executes the computer-readable instructions to implement the steps of the artificial intelligence-based strategy generation method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of the artificial intelligence-based strategy generation method as described in any one of claims 1 to 7.

Images