Insurance product personalized recommendation system based on artificial intelligence

Abstract

The invention relates to an insurance product personalized recommendation system based on artificial intelligence, and the system comprises a customer portrait module which extracts implicit features and dominant features in portrait data through a preset deep neural network technology, and constructs a customer portrait knowledge graph based on the combination of the implicit features and the dominant features; the insurance product knowledge base module is used for collecting various insurance product data, extracting implicit attributes and dominant attributes of insurance products in the insurance product data by utilizing a preset natural language processing technology, and combining the implicit attributes with the dominant attributes to construct an insurance product knowledge graph; the disease diagnosis module is used for acquiring the doctor seeing information of the customer and extracting the disease diagnosis information of the customer based on the doctor seeing information so as to predict the future health condition and the disease risk level of the customer according to the disease diagnosis information; and the insurance recommendation module is used for generating an insurance recommendation list of the customer according to the customer portrait knowledge graph, the insurance product knowledge graph, the future health condition of the customer and the disease risk level.

Description

Technical Field

[0001] The present invention relates to the field of insurance technology, and in particular to an insurance product personalized recommendation system based on artificial intelligence. Background Art

[0002] In the related art, an insurance product recommendation system based on collaborative filtering is proposed. The system uses the customer's historical purchase records and evaluation data to calculate the similarity between customers or insurance products, and recommends similar or favorite insurance products to target customers based on the purchase or evaluation of similar customers. However, the disadvantage of this system is that it requires a large amount of user behavior data, lacks an effective recommendation method for new customers or new products, and cannot consider other characteristics and needs of customers, as well as other attributes and matching of insurance products.

[0003] There are product implementation cases on the market, such as the Smart Insurance Consultant APP, which combines the family information and protection needs entered by the user and customizes a unique family insurance configuration plan for the user through a specific smart insurance recommendation algorithm. By setting extremely simple questions, key information about the customer is obtained in the shortest possible time, thereby achieving demand analysis.

[0004] However, smart assistants like this are not truly customer-centric. The questions they ask are extremely simple and they do not provide a comprehensive understanding of the customers. They are unable to understand the customers’ real needs based on just a few simple questions, so their insurance product recommendations are inaccurate.

[0005] In summary, the insurance recommendation system in the prior art lacks a comprehensive and in-depth understanding of customers, lacks a rich and accurate description of insurance products, lacks the processing of complex and dynamic customer needs, and lacks optimization of insurance product matching and recommendations. Summary of the invention

[0006] The present invention provides an artificial intelligence-based personalized insurance product recommendation system to solve the problems of the insurance recommendation system in the prior art, such as lack of comprehensive and in-depth understanding of customers, lack of rich and accurate description of insurance products, lack of processing of complex and dynamic customer needs, and lack of optimization of insurance product matching and recommendation.

[0007] An embodiment of the present invention provides an artificial intelligence-based personalized recommendation system for insurance products, including: a customer portrait module, which is used to collect customer portrait data, and use a preset deep neural network technology to extract implicit features and explicit features in the portrait data, and construct a customer portrait knowledge graph based on the implicit features and the explicit features; an insurance product knowledge base module, which is used to collect multiple insurance product data, and use a preset natural language processing technology to extract implicit attributes and explicit attributes of insurance products in the insurance product data, and combine the implicit attributes with the explicit attributes to construct an insurance product knowledge graph; a disease diagnosis module, which is used to obtain the customer's medical information, and extract the customer's disease diagnosis information based on the medical information, so as to predict the customer's future health status and disease risk level based on the disease diagnosis information; an insurance recommendation module, which is used to generate an insurance recommendation list for the customer based on the customer portrait knowledge graph, the insurance product knowledge graph, the customer's future health status and disease risk level.

[0008] Optionally, the insurance recommendation module includes: an insurance demand analysis unit, which is used to analyze the customer's risk prevention and wealth management needs based on the customer portrait knowledge graph and the disease diagnosis information using a preset deep learning technology, and convert the risk prevention and wealth management needs into customer demand evaluation indicators, so as to dynamically analyze the customer's insurance product needs based on the customer demand evaluation indicators; an insurance product matching unit, which is used to calculate the matching degree between each insurance product in the insurance product knowledge graph and the insurance product demand using a preset multi-objective optimization technology, and sort and filter the insurance products in the insurance product knowledge graph according to the matching degree to obtain the customer's insurance recommendation list, wherein the matching degree calculation formula is:

[0009] Matching Score=w1×(1-Euclidean Distance)+w2×Cosine Similarity;

[0010] Among them, w1 is the weight of Euclidean distance, w2 is the weight of cosine similarity, Cosime Similarity is to maximize cosine similarity, and Euclidean Distance is Euclidean distance.

[0011] Optionally, the insurance recommendation module also includes: a product recommendation unit, which is used to use preset explainable machine learning technology to perform in-depth analysis on the insurance recommendation list and generate corresponding explanations and suggestions for each insurance product; and a recommendation feedback unit, which is used to receive the customer's evaluation and feedback on the insurance recommendation list.

[0012] Optionally, the insurance demand analysis unit includes: an adjustment subunit, configured to obtain the budget information of the customer, and adjust and optimize the insurance recommendation list according to the budget information.

[0013] Optionally, the insurance product matching unit includes: a selection subunit, configured to obtain the insurance product matching target selected by the customer, and generate the insurance recommendation list according to the insurance product matching target.

[0014] Optionally, the artificial intelligence-based personalized insurance product recommendation system also includes: an interactive interface module, which is used to collect the customer's personal information data, preferred product data and product demand data, and provide the customer with an interface for viewing or selecting insurance products.

[0015] Optionally, the interactive interface module includes: a voice input unit for receiving personal information data, preferred product data and product demand data sent by the customer through voice, and converting the personal information data, the preferred product data and the product demand data into text and displaying them on the interface.

[0016] Optionally, the customer portrait module includes: a user feedback unit, used to receive the customer's evaluation and modification of the customer portrait knowledge graph.

[0017] Optionally, the insurance product knowledge base module includes: a product update unit, configured to obtain the latest insurance product data from a third party according to a preset period, and update the insurance product knowledge graph based on the latest insurance product data.

[0018] Optionally, the disease diagnosis module includes: a disease prediction unit, which is used to predict the customer's future health status and disease risk level based on the disease diagnosis information using preset machine learning technology, and generate prevention recommendations for the customer's specific situation based on the future health status and the disease risk level.

[0019] In the above implementation, a customer portrait module is used to collect customer portrait data, and a preset deep neural network technology is used to extract implicit features and explicit features in the portrait data, and a customer portrait knowledge graph is constructed based on the implicit features and explicit features. An insurance product knowledge base module is used to collect a variety of insurance product data, and a preset natural language processing technology is used to extract implicit attributes and explicit attributes of insurance products in the insurance product data, and the implicit attributes are combined with the explicit attributes to construct an insurance product knowledge graph. A disease diagnosis module is used to obtain a customer's medical information, and the customer's disease diagnosis information is extracted based on the medical information to predict the customer's future health status and disease risk level based on the disease diagnosis information, and an insurance recommendation module is used to generate a customer's insurance recommendation list based on the customer portrait knowledge graph, the insurance product knowledge graph, the customer's future health status and disease risk level. As a result, the problems of the insurance recommendation system in the prior art, such as the lack of comprehensive and in-depth understanding of customers, lack of rich and accurate description of insurance products, lack of processing of complex and dynamic customer needs, and lack of optimization of insurance product matching and recommendation, are solved. The system can fully and deeply understand the characteristics of customers, richly and accurately describe the attributes of insurance products, automatically analyze the health status and risk level of customers, and convert them into quantifiable evaluation indicators, analyze customer needs in a complex and dynamic manner, optimize and optimally match and recommend insurance products, and provide corresponding explanations and suggestions.

[0020] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0022] Figure 1 A block diagram of an insurance product personalized recommendation system based on artificial intelligence provided according to an embodiment of the present invention;

[0023] Figure 2 A schematic diagram of an insurance product personalized recommendation system based on artificial intelligence according to an embodiment of the present invention;

[0024] Figure 3 The present invention is a flowchart of a personalized insurance product recommendation method based on artificial intelligence according to an embodiment of the present invention. Description of the drawings:

[0026] 10-Personalized recommendation system for insurance products based on artificial intelligence; 100-Customer portrait module; 200-Insurance product knowledge base module; 300-Disease diagnosis module; 400-Insurance recommendation module. DETAILED DESCRIPTION

[0027] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limiting the present invention.

[0028] The following describes an artificial intelligence-based personalized recommendation system for insurance products according to an embodiment of the present invention with reference to the accompanying drawings. In view of the problem that the insurance recommendation system in the prior art mentioned in the above background technology lacks a comprehensive and in-depth understanding of customers, lacks a rich and accurate description of insurance products, lacks the processing of complex and dynamic customer needs, and lacks the optimization of insurance product matching and recommendation, the present invention provides an artificial intelligence-based personalized recommendation system for insurance products, in which a customer portrait module is used to collect customer portrait data, and a preset deep neural network technology is used to extract implicit features and explicit features in the portrait data, and a customer portrait knowledge graph is constructed based on the implicit features and the explicit features, an insurance product knowledge base module is used to collect a variety of insurance product data, and a preset natural language processing technology is used to extract implicit attributes and explicit attributes of insurance products in the insurance product data, and the implicit attributes and the explicit attributes are combined to construct an insurance product knowledge graph, a disease diagnosis module is used to obtain customer medical information, and the customer's disease diagnosis information is extracted based on the medical information, so as to predict the customer's future health status and disease risk level according to the disease diagnosis information, and an insurance recommendation module is used to generate a customer's insurance recommendation list according to the customer portrait knowledge graph, the insurance product knowledge graph, the customer's future health status and the disease risk level. As a result, the problems of the insurance recommendation system in the prior art, such as the lack of comprehensive and in-depth understanding of customers, lack of rich and accurate description of insurance products, lack of processing of complex and dynamic customer needs, and lack of optimization of insurance product matching and recommendation, are solved. The system can fully and deeply understand the characteristics of customers, richly and accurately describe the attributes of insurance products, automatically analyze the health status and risk level of customers, and convert them into quantifiable evaluation indicators, analyze customer needs in a complex and dynamic manner, optimize and optimally match and recommend insurance products, and provide corresponding explanations and suggestions.

[0029] The insurance product recommendation system in the prior art only uses the customer's historical purchase records and evaluation data to calculate the similarity between customers or insurance products, and recommends similar or favorite insurance products to the target customer based on the purchase or evaluation of similar customers. This method has the following disadvantages:

[0030] (1) Lack of comprehensive and in-depth understanding of the client. This method only makes recommendations based on the basic information provided by the client (such as age, gender, income, etc.), ignoring other important characteristics of the client (such as family status, health status, lifestyle habits, risk preferences, etc.). This results in the recommendation results being lacking in specificity and personalization. Because different clients may have different risk prevention and wealth management goals, relying solely on historical data cannot reflect the client’s true needs and preferences.

[0031] (2) Lack of rich and accurate description of insurance products. This method only makes recommendations based on simple attributes such as the name, type, and price of insurance products, ignoring other important attributes of insurance products (such as coverage, liability limit, deductible, compensation conditions, etc.). This results in a lack of credibility and transparency in the recommendation results, because different insurance products may have different coverage content and liability limits. Relying solely on simple attributes cannot reflect the true characteristics and advantages and disadvantages of insurance products.

[0032] (3) Lack of processing of the complexity and dynamics of customer needs. This method only makes recommendations based on the customer's single or static needs, ignoring the customer's multiple or dynamic needs. This leads to a lack of flexibility and adaptability in the recommendation results. Because customer needs may change with time, environment, events and other factors, relying solely on historical data cannot reflect the customer's current needs and changes.

[0033] (4) Lack of optimization for insurance product matching and recommendation. This method matches and recommends insurance products based only on simple rules or models, ignoring the relationships and influences between insurance products. This results in a lack of optimization and optimality in the recommendation results. This is because different insurance products may have different combination effects and competitive effects. Simply relying on similarity cannot reflect the coordination and competitiveness between insurance products.

[0034] Specifically, Figure 1 A flowchart of an insurance product personalized recommendation system based on artificial intelligence provided by an embodiment of the present invention.

[0035] like Figure 1 As shown, the insurance product personalized recommendation system based on artificial intelligence includes: a customer portrait module 100, an insurance product knowledge base module 200, a disease diagnosis module 300 and an insurance recommendation module 400, wherein:

[0036] The customer portrait module 100 is used to collect the customer's portrait data, and use the preset deep neural network technology to extract the implicit features and explicit features in the portrait data, and construct the customer portrait knowledge graph based on the implicit features and the explicit features; the insurance product knowledge base module 200 is used to collect a variety of insurance product data, and use the preset natural language processing technology to extract the implicit attributes and explicit attributes of the insurance products in the insurance product data, and combine the implicit attributes with the explicit attributes to construct the insurance product knowledge graph; the disease diagnosis module 300 is used to obtain the customer's medical information, and extract the customer's disease diagnosis information based on the medical information, so as to predict the customer's future health status and disease risk level based on the disease diagnosis information; the insurance recommendation module 400 is used to generate a customer's insurance recommendation list based on the customer portrait knowledge graph, the insurance product knowledge graph, the customer's future health status and disease risk level.

[0037] Specifically, Figure 2 As shown, the customer portrait module 100 is responsible for collecting and processing the portrait data input or uploaded by the customer, and using deep neural network technology to automatically learn the customer's implicit characteristics (such as risk preference, wealth goals, etc.), and combine it with the customer's explicit characteristics (such as age, gender, income, etc.) to build a customer portrait knowledge graph to fully and deeply understand the customer's characteristics. This module can be implemented using different deep neural network technologies, such as CNN, RNN, Transformer, etc. These technologies have different advantages and disadvantages. For example, CNN is suitable for processing image data, RNN is suitable for processing time series data, Transformer is suitable for processing natural language data, etc. Different technical implementation methods are selected according to different data types and scenarios.

[0038] The insurance product knowledge base module 200 is a core module for describing the attributes of insurance products. It is responsible for collecting and processing data of insurance products of various types and prices on the market, and automatically extracting implicit attributes of insurance products (such as coverage, liability limits, deductibles, compensation conditions, etc.) using natural language processing technology, and combining them with explicit attributes of insurance products (such as name, type, price, etc.) to build an insurance product knowledge graph to enrich and accurately describe the attributes of insurance products. This module is implemented using different natural language processing technologies, such as BERT, GPT-3, XLNet, etc. These technologies can achieve the same function, that is, automatically extracting implicit attributes of insurance products and combining them with explicit attributes of insurance products. There are different advantages and disadvantages between these technologies. For example, BERT is suitable for processing bidirectional contexts, GPT-3 is suitable for processing generative tasks, and XLNet is suitable for processing long texts. Different technical implementation methods are selected according to different task types and difficulties.

[0039] The disease diagnosis module 300 is a core module for analyzing the customer's health status and risk level. It is responsible for extracting relevant disease diagnosis information (such as disease name, severity, treatment plan, etc.) from the physical examination report or medical record uploaded by the customer, and converting it into quantifiable evaluation indicators (such as risk level, protection needs, etc.) to accurately and personally analyze the customer's health status and risk level. This module is implemented using different medical knowledge graphs and medical natural language processing technologies, such as BioBERT, MedCAT, SciBERT, etc. These technologies can achieve the same function, that is, extracting relevant disease diagnosis information from the physical examination report or medical record uploaded by the customer, and converting it into quantifiable evaluation indicators. These technologies have different advantages and disadvantages. For example, BioBERT is suitable for processing texts in the biomedical field, MedCAT is suitable for processing texts in the clinical field, SciBERT is suitable for processing texts in the scientific field, etc. Different technical implementation methods are selected according to different text fields and quality.

[0040] The insurance product recommendation module 400 is the core module of the system for recommending insurance products. It is responsible for generating a list of insurance product recommendations based on the sorted and filtered results, and providing corresponding explanations and suggestions (such as reasons for recommendation, analysis of advantages and disadvantages, risk point prompts, precautions, etc.) to increase the interpretability and credibility of the recommendations. This module can be implemented using explainable machine learning technologies such as LIME and SHAP.

[0041] Optionally, in some embodiments, the insurance recommendation module 400 includes: an insurance demand analysis unit, which is used to analyze the customer's risk prevention and wealth management needs based on the customer portrait knowledge graph and disease diagnosis information using a preset deep learning technology, and convert the risk prevention and wealth management needs into customer demand evaluation indicators, so as to dynamically analyze the customer's insurance product needs based on the customer demand evaluation indicators; an insurance product matching unit, which is used to calculate the matching degree between each insurance product in the insurance product knowledge graph and the insurance product demand using a preset multi-objective optimization technology, and sort and filter the insurance products in the insurance product knowledge graph according to the matching degree to obtain the customer's insurance recommendation list, wherein the matching degree calculation formula is:

[0042] Matching Score=w1×(1-Euclidean Distance)+w2×Cosine SImilarity;

[0043] Among them, w1 is the weight of Euclidean distance, w2 is the weight of cosine similarity, Cosine Similarity is to maximize cosine similarity, and Euclidean Distance is Euclidean distance.

[0044] Specifically, the insurance demand analysis unit is responsible for automatically analyzing the customer's risk prevention and wealth management needs based on the customer portrait knowledge graph and disease diagnosis information using deep learning technology, and converting them into quantifiable evaluation indicators (such as life insurance needs, medical insurance needs, accident insurance needs, property insurance needs, etc.) to analyze customer needs in a complex and dynamic manner. This module can be implemented using different deep learning technologies, such as LSTM, Transformer, GRU, etc. These technologies can all achieve the same function, that is, automatically analyzing the customer's risk prevention and wealth management needs based on the customer portrait knowledge graph and disease diagnosis information, and converting them into quantifiable evaluation indicators. These technologies have different advantages and disadvantages. For example, LSTM is suitable for dealing with long-term dependency problems, Transformer is suitable for dealing with parallel computing problems, GRU is suitable for dealing with simple and fast problems, etc. Different technical implementation methods are selected according to different computing resources and efficiency.

[0045] The insurance product matching unit is responsible for automatically calculating the matching degree between each insurance product and customer demand based on the customer demand evaluation index and the insurance product attribute evaluation index, and sorting and filtering the insurance products according to the matching degree, so as to optimize and optimally match the insurance products. This module is implemented using different multi-objective optimization technologies, such as NSGA-II, MOEA / D, SPEA2, etc. These technologies can achieve the same function, that is, automatically calculating the matching degree between each insurance product and customer demand based on the customer demand evaluation index and the insurance product attribute evaluation index, and sorting and filtering the insurance products according to the matching degree. These technologies have different advantages and disadvantages. For example, NSGA-II is suitable for dealing with diversity problems, MOEA / D is suitable for dealing with decomposition problems, SPEA2 is suitable for dealing with intensity problems, etc. Different technical implementation methods are selected according to different optimization goals and constraints.

[0046] In order to enable those skilled in the art to further understand the implementation method of the insurance product recommendation module in the embodiment of the present invention, the following is a detailed description in conjunction with a specific embodiment:

[0047] Name: Zhang San

[0048] Gender: Male

[0049] Risk appetite: Medium (value expressed as 0.5)

[0050] Investment preference: long-term investment (value expressed as 0.8)

[0051] Health index: High (value expressed as 0.7)

[0052] Liver-related disease susceptibility index: high (the numerical value is 0.7)

[0053] Insurance Product A:

[0054] property:

[0055] Investment period: long term (value expressed as 0.8)

[0056] Risk level: Medium (value expressed as 0.5)

[0057] Premium: Moderate (value expressed as 0.6)

[0058] Liver-related protection level: Medium (value expressed as 0.5)

[0059] Insurance Product B:

[0060] property:

[0061] Investment period: Medium term (value expressed as 0.4)

[0062] Risk level: High (value expressed as 0.8)

[0063] Premium: High (value expressed as 0.7)

[0064] Liver-related protection level: High (the numerical value is 0.8)

[0065] Steps to implement the program:

[0066] 1. Define customer needs and insurance product attribute evaluation indicators

[0067] Customer demand evaluation indicators: risk preference, investment period preference, health index, and liver-related disease susceptibility index.

[0068] Insurance product attribute evaluation indicators: investment period, risk level, premium, and liver-related protection level.

[0069] 2. Mathematical modeling

[0070] Zhang San’s demand vector: [0.5, 0.8, 0.7, 0.7]

[0071] Attribute vector of insurance product A: [0.8, 0.5, 0.6, 0.5]

[0072] Attribute vector of insurance product B: [0.4, 0.8, 0.7, 0.8]

[0073] 3. Objective function definition

[0074] The objective function may include minimizing the distance to the customer demand vector while maximizing the similarity to the customer demand vector.

[0075] Goal 1: Minimize Euclidean distance

[0076] Calculate the Euclidean distance between the customer demand vector and the insurance product attribute vector. The smaller the Euclidean distance, the closer the product attributes are to the customer demand.

[0077]

[0078] Objective 2: Maximize cosine similarity

[0079] Calculate the cosine similarity between the customer demand vector and the insurance product attribute vector. The higher the cosine similarity, the closer the product attributes are to the customer demand.

[0080]

[0081] Where C·P represents the dot product of the vectors, and ||C|| and ||P|| represent the modulus of the vectors.

[0082] 4. Multi-objective optimization calculation

[0083] Use a multi-objective optimization algorithm, such as NSGA-II, to calculate the matching degree between insurance product A and insurance product B and Zhang San's needs. The following is an example:

[0084] The Euclidean distance between product A and customer demand:

[0085]

[0086] The Euclidean distance between product B and customer demand:

[0087]

[0088] Cosine similarity between product A and customer demand:

[0089]

[0090] Cosine similarity between product B and customer demand:

[0091]

[0092] Combine the weights to calculate the final matching degree:

[0093] Assuming that the weight of Euclidean distance is w1=0.4 and the weight of cosine similarity is w2=0.6, the final matching degree calculation formula is:

[0094] Matching Score=w1×(1-Euclidean Distance)+w2×Cosine Similarity

[0095] Product A's matching degree:

[0096] Matching ScoreA =0.4×(1-0.48)+0.6×0.937≈0.4×0.52+0.6×0.937≈0.208+0.5622≈0.77

[0097] Matching Score B =0.4×(1-0.14)+0.6×0.906≈0.4×0.86+0.6×0.906≈0.344+0.5436≈0.8876

[0098] Product B's matching degree:

[0099] Product A’s matching degree: 0.77

[0100] Product B's matching degree: 0.8876

[0101] 5. Calculate the matching degree

[0102] The algorithm output is:

[0103] Insurance product A’s matching score: 0.77

[0104] Matching score of insurance product B: 0.8876

[0105] 6. Sorting and filtering

[0106] Taking into account Zhang San's liver-related disease susceptibility index, insurance product B scores higher in liver-related protection, so B will be recommended to Zhang San first.

[0107] Optionally, in some embodiments, the insurance recommendation module 400 further includes: a product recommendation unit, which is used to use a preset explainable machine learning technology to perform an in-depth analysis of the insurance recommendation list and generate corresponding explanations and suggestions for each insurance product; and a recommendation feedback unit, which is used to receive customers' evaluations and feedback on the insurance recommendation list.

[0108] In the insurance product recommendation module 400, a recommendation feedback unit and a product suggestion unit are added. The product suggestion unit can be implemented using interpretable machine learning technologies such as LIME and SHAP to conduct in-depth analysis of the insurance recommendation list and provide corresponding explanations and suggestions (such as reasons for recommendation, analysis of advantages and disadvantages, risk point prompts, precautions, etc.).

[0109] The recommendation feedback unit allows customers to evaluate and provide feedback on the insurance products recommended by the system, such as giving likes or dislikes, purchasing or rejecting, commenting or making suggestions, etc. This can improve the system's recommendation effectiveness and continuous improvement capabilities.

[0110] Optionally, in some embodiments, the insurance demand analysis unit includes: an adjustment subunit configured to obtain the budget information of the customer and adjust and optimize the insurance recommendation list according to the budget information.

[0111] In the insurance demand analysis unit, an adjustment subunit is added, enabling the customer to adjust and optimize the insurance demand evaluation indicators generated by the system based on their actual situation and budget, such as increasing or decreasing certain types or amounts of insurance demands. This can improve the customer's demand satisfaction and right to independent choice.

[0112] Optionally, in some embodiments, the insurance product matching unit includes: a selection subunit configured to obtain the insurance product matching target selected by the customer and generate an insurance recommendation list according to the insurance product matching target.

[0113] In the insurance product matching unit, a selection subunit is added, allowing the customer to select different matching strategies according to their preferred insurance product matching targets, such as maximizing protection, minimizing cost, optimizing the combination, etc. This can improve the customer's matching satisfaction and personalized customization ability.

[0114] Optionally, in some embodiments, it further includes: an interaction interface module configured to collect the personal information data, preferred product data, and product demand data of the customer and provide an interface for the customer to view or select insurance products.

[0115] Specifically, the interaction interface module is the front-end interface for the system to interact with the customer, responsible for providing a friendly and convenient user interface that enables the customer to input or upload their personal information, preferences, demands, etc., and view or select the insurance products recommended by the system. This module can be implemented in the form of a web page, APP, WeChat mini-program, etc., allowing the customer to conveniently use the system's services and improving the customer's experience and satisfaction.

[0116] For example, the interaction interface is in the form of a questionnaire. First, it asks the user to fill in some personal information, such as name, age, etc. These are fixed fields for storage in the database, and the data format is the JSON object collected by the page, such as: {name:'张三',age:18,sex:'man',address:'北京市'}.

[0117] After filling in the basic information, users can upload some physical examination reports or medical records. At this time, ORC needs to be used to parse the uploaded pictures or PDF files, and then the parsed content is mapped into the above type of JSON objects, such as: the systolic pressure is 116 and the diastolic pressure is 66 in the physical examination report, which is finally parsed into {systolicPressure:116,diastolicPressure:66}. These conversions are standardized and require a large number of related fields to be defined in advance. For example, systolic pressure needs to be mapped to the fixed systolicPressure field.

[0118] Optionally, in some embodiments, the interactive interface module includes: a voice input unit for receiving personal information data, preferred product data and product demand data sent by the customer through voice, and converting the personal information data, preferred product data and product demand data into text and displaying it on the interface.

[0119] Specifically, a voice input unit is added to the interactive interface module, allowing customers to choose to say their basic information, such as age, gender, physical condition, etc. The backend will then receive these voice commands and convert them into text and display them on the page for customers to edit again, which can improve customer convenience and satisfaction.

[0120] Optionally, in some embodiments, the customer portrait module 100 includes: a user feedback unit for receiving customers' evaluations and modifications to the customer portrait knowledge graph.

[0121] A user feedback unit has been added to the customer portrait module, allowing customers to evaluate and modify the customer portrait knowledge graph generated by the system, such as confirming or denying certain characteristics, preferences, needs, etc., which can improve the accuracy and reliability of the customer portrait.

[0122] Optionally, in some embodiments, the insurance product knowledge base module 200 includes: a product update unit, configured to obtain the latest insurance product data from a third party according to a preset period, and update the insurance product knowledge graph based on the latest insurance product data.

[0123] In the insurance product knowledge base module 200, a product update unit is added, so that the system can regularly obtain the latest insurance product data from insurance companies or third-party platforms and update the insurance product knowledge graph, which can improve the timeliness and completeness of insurance products.

[0124] Optionally, in some embodiments, the disease diagnosis module 300 includes: a disease prediction unit, which is used to predict the client's future health status and disease risk level based on the disease diagnosis information using preset machine learning technology, and generate prevention recommendations for the client's specific situation based on the future health status and disease risk level.

[0125] In the disease diagnosis module 300, a disease prediction unit is added, so that the system can use machine learning technology to predict the diseases that customers may suffer from in the future based on factors such as the customer's age, gender, family history, and living habits, and give corresponding prevention suggestions, which can improve the customer's health awareness and risk avoidance ability.

[0126] In order to enable those skilled in the art to further understand the personalized recommendation system for insurance products based on artificial intelligence according to the embodiment of the present invention, the following is a detailed description in conjunction with specific embodiments. Figure 3 shown.

[0127] Assume that there is a 30-year-old male customer named Zhang San who wants to buy an insurance product suitable for him. He enters his basic information (such as name, age, gender, income, risk preference, hobbies, physical condition, etc.) through the interactive interface module provided by the present invention, and uploads his most recent physical examination report and historical medical diagnosis records.

[0128] The customer portrait module collects and processes the data input and uploaded by Zhang San, and uses deep neural network technology to automatically learn Zhang San's implicit features (such as risk preference, wealth goals, etc.), and combines them with Zhang San's explicit features to build Zhang San's customer portrait knowledge graph. For example, Zhang San is a person with career development potential, pursuing high-quality life, focusing on health prevention, and planning for the future. He has a family history of cancer and needs to pay attention to cancer-related insurance configuration;

[0129] The disease diagnosis module extracts relevant disease diagnosis information from the physical examination report uploaded by Zhang San, such as Zhang San suffers from thyroid nodule disease, which is a medium-risk level, and converts it into quantifiable evaluation indicators, such as Zhang San’s demand for coverage is 0.8, the demand for liability limit is 0.6, the demand for deductible is 0.4, the demand for compensation conditions is 0.9, etc.

[0130] The insurance product knowledge base module collects and processes data on insurance products of various types and prices on the market, and uses natural language processing technology to automatically extract the implicit attributes of insurance products (such as coverage, liability limits, deductibles, compensation conditions, etc.), and combines them with the explicit attributes of insurance products to construct an insurance product knowledge graph.

[0131] The insurance demand analysis module uses deep learning technology to automatically analyze Zhang San’s risk prevention and wealth management needs based on his customer portrait knowledge graph and disease diagnosis information, and converts them into quantifiable evaluation indicators. For example, Zhang San’s demand for life insurance is 0.5, his demand for medical insurance is 0.8, his demand for accident insurance is 0.6, and his demand for critical illness insurance is 0.7.

[0132] The insurance product matching module automatically calculates the matching degree between each insurance product and Zhang San's needs based on Zhang San's demand evaluation indicators and insurance product attribute evaluation indicators, and uses multi-objective optimization technology to sort and filter the insurance products according to the matching degree.

[0133] The insurance product recommendation module generates a list of recommended insurance products based on the sorted and filtered results, and provides corresponding explanations and suggestions. For example, the first insurance product in the recommended list is product C, and its explanation and suggestions are as follows:

[0134] Reason for recommendation: Product C is a professional critical illness insurance product that can provide you with protection against 100 major diseases, including the thyroid disease you suffer from. Product C has a match degree with your needs as high as 0.95, making it one of the most suitable insurance products for you.

[0135] The following is a brief analysis of this product. If you want to view more detailed analysis at the clause level, you can click to download the report we generated for you to view the details.

[0136] Advantages and disadvantages analysis: The advantages of product C are that it covers a wide range of major diseases on the market, with a liability limit of up to 500,000 yuan, which can provide you with sufficient financial support, and the insurance conditions are loose. Your thyroid nodules can also be insured normally. After simulating the insurance company's intelligent underwriting for you, the pass rate is as high as 99%. The compensation conditions of this product are simple and clear. As long as a major disease is diagnosed, it can be paid in one lump sum. Moderate and mild diseases can also be compensated up to 6 times without grouping.

[0137] The disadvantage of Product C is its higher price. For example, if you choose a policy with a coverage of 500,000 yuan and lifetime coverage, you will need to pay an annual premium of 7,000 yuan, which may be a burden for your income level. You can choose a more suitable combination of terms at your discretion.

[0138] Risk Tips: Although Product C can cover your thyroid disease, there is an age limit for insurance. If you are over 35 years old, Product C will not accept your insurance application. Please read the terms carefully before purchasing and make a decision based on your actual situation.

[0139] Notes and suggestions: If you decide to purchase product C, please follow the following process: click the purchase button on the interactive interface module, fill in the relevant information, and submit the order; after receiving the underwriting notice sent by the insurance company, follow the instructions to perform the underwriting operation and pay the premium; after receiving the electronic insurance policy sent by the insurance company, save the electronic insurance policy and print out the paper insurance policy if necessary; after an insurance accident occurs, contact the insurance company in a timely manner and submit relevant supporting documents in accordance with the claims process;

[0140] Therefore, the beneficial effects of the present invention are as follows:

[0141] (1) Comprehensively and deeply understand customer characteristics, build customer profiles, and improve the targeting and personalization of recommendations;

[0142] (2) Enrich and accurately describe the attributes of insurance products, build an insurance product knowledge base, and improve the credibility and transparency of recommendations;

[0143] (3) Based on the physical examination report or medical record uploaded by the customer, the company uses disease diagnosis technology to automatically analyze the customer's health status and risk level and convert it into quantifiable evaluation indicators;

[0144] (4) Analyze customer needs in a complex and dynamic manner, build insurance demand models, and improve the flexibility and adaptability of recommendations;

[0145] (5) Optimize and optimally match and recommend insurance products, build insurance product matching and recommendation models, and provide corresponding explanations and suggestions.

[0146] According to the artificial intelligence-based personalized recommendation system for insurance products proposed in an embodiment of the present invention, a customer portrait module is used to collect customer portrait data, and a preset deep neural network technology is used to extract implicit features and explicit features in the portrait data, and a customer portrait knowledge graph is constructed based on the combination of implicit features and explicit features. An insurance product knowledge base module is used to collect multiple insurance product data, and a preset natural language processing technology is used to extract implicit attributes and explicit attributes of insurance products in the insurance product data, and the implicit attributes are combined with the explicit attributes to construct an insurance product knowledge graph. A disease diagnosis module is used to obtain a customer's medical information, and the customer's disease diagnosis information is extracted based on the medical information, so as to predict the customer's future health status and disease risk level based on the disease diagnosis information, and an insurance recommendation list for the customer is generated through the insurance recommendation module based on the customer portrait knowledge graph, the insurance product knowledge graph, the customer's future health status and disease risk level. As a result, the problems of the insurance recommendation system in the prior art, such as the lack of comprehensive and in-depth understanding of customers, lack of rich and accurate description of insurance products, lack of processing of complex and dynamic customer needs, and lack of optimization of insurance product matching and recommendation, are solved. The system can fully and deeply understand the characteristics of customers, richly and accurately describe the attributes of insurance products, automatically analyze the health status and risk level of customers, and convert them into quantifiable evaluation indicators, analyze customer needs in a complex and dynamic manner, optimize and optimally match and recommend insurance products, and provide corresponding explanations and suggestions.

[0147] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or N embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0148] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "N" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

Claims

1. An insurance product personalized recommendation system based on artificial intelligence, characterized in that: include: The customer portrait module is used to collect customer portrait data, extract implicit features and explicit features in the portrait data using a preset deep neural network technology, and construct a customer portrait knowledge graph based on the implicit features and the explicit features. An insurance product knowledge base module is used to collect data on various insurance products, and use a preset natural language processing technology to extract implicit attributes and explicit attributes of insurance products in the insurance product data, and combine the implicit attributes with the explicit attributes to construct an insurance product knowledge graph; A disease diagnosis module, used to obtain the customer's medical information and extract the customer's disease diagnosis information based on the medical information, so as to predict the customer's future health status and disease risk level according to the disease diagnosis information; The insurance recommendation module is used to generate an insurance recommendation list for the customer based on the customer portrait knowledge graph, the insurance product knowledge graph, the customer's future health status and disease risk level.

2. The system according to claim 1, characterized in that The insurance recommendation module comprises: An insurance demand analysis unit, configured to analyze the risk prevention and wealth management needs of the customer based on the customer portrait knowledge graph and the disease diagnosis information using a preset deep learning technology, and convert the risk prevention and wealth management needs into customer demand evaluation indicators, so as to dynamically analyze the customer's insurance product needs based on the customer demand evaluation indicators; The insurance product matching unit is used to calculate the matching degree between each insurance product in the insurance product knowledge graph and the insurance product demand by using a preset multi-objective optimization technology, and sort and filter the insurance products in the insurance product knowledge graph according to the matching degree to obtain the insurance recommendation list of the customer, wherein the matching degree is calculated as follows: Matching Score=w1×(1-Euclidean Distance)+w2×Cosine Similarity; Among them, w1 is the weight of Euclidean distance, w2 is the weight of cosine similarity, Cosine Similarity is to maximize cosine similarity, and Euclidean Distance is Euclidean distance.

3. The system according to claim 2, characterized in that The insurance recommendation module further includes: A product recommendation unit, configured to use a preset interpretable machine learning technology to perform an in-depth analysis on the insurance recommendation list and generate corresponding explanations and recommendations for each insurance product; The recommendation feedback unit is used to receive the customer's evaluation and feedback on the insurance recommendation list.

4. The system according to claim 2, characterized in that The insurance demand analysis unit comprises: The adjustment subunit is used to obtain the budget information of the customer and adjust and optimize the insurance recommendation list according to the budget information.

5. The system according to claim 2, characterized in that The insurance product matching unit includes: The selection subunit is used to obtain the insurance product matching target selected by the customer and generate the insurance recommendation list according to the insurance product matching target.

6. The system according to claim 1, characterized in that Also includes: The interactive interface module is used to collect the customer's personal information data, preferred product data and product demand data, and provide the customer with an interface for viewing or selecting insurance products.

7. The system according to claim 6, characterized in that The interactive interface module comprises: The voice input unit is used to receive the personal information data, preferred product data and product demand data sent by the customer through voice, and convert the personal information data, the preferred product data and the product demand data into text and display them on the interface.

8. The system according to claim 1, characterized in that The customer portrait module includes: A user feedback unit is used to receive the customer's evaluation and modification of the customer portrait knowledge graph.

9. The system according to claim 1, characterized in that The insurance product knowledge base module includes: A product update unit is used to obtain the latest insurance product data from a third party according to a preset period, and update the insurance product knowledge graph based on the latest insurance product data.

10. The system according to claim 1, characterized in that The disease diagnosis module comprises: A disease prediction unit is used to predict the customer's future health status and disease risk level based on the disease diagnosis information using a preset machine learning technology, and generate prevention suggestions for the customer's specific situation based on the future health status and the disease risk level.

Images