Insurance product multi-dimensional comparative analysis method and system based on big data

Abstract

The invention relates to the technical field of insurance product comparative analysis, in particular to an insurance product multi-dimensional comparative analysis method and system based on big data, and the method comprises the steps: calculating insurance type matching scores for users with consultation interaction behaviors through a keyword-insurance type weight matrix, and calculating insurance type matching scores for users without consultation behaviors through a keyword-insurance type weight matrix; matching neighbor user group preferences to determine an inclined insurance type so as to pre-judge the inclined insurance type of the user; according to a pre-judged user tendency insurance type, insurance product attributes of the insurance type are divided into common attributes and specific attributes, the attributes are converted into horizontal and vertical coordinate labeling positions, a personalized demand matching area is generated in combination with user demands, an insurance product recommendation set is generated, and maximum characteristics and defects are labeled, so that accurate comparison and recommendation of insurance products are realized. And the user decision-making efficiency is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of insurance product comparative analysis, more particularly, it relates to a big data-based insurance product multi-dimensional comparative analysis method and system. BACKGROUND

[0002] Insurance product analysis is an important research direction in the intersection field of financial technology and big data application, and its core goal is to integrate insurance product information through technical means, help users quickly filter out products that meet their own needs, and reduce decision-making difficulty. With the rapid development of the insurance market, product categories are increasingly diverse, and health insurance, car insurance, property insurance and other sub-insurance products are constantly being updated, and users' demand for accurate and personalized product recommendations is increasingly urgent.

[0003] The prior art has significant limitations in insurance product analysis and recommendation. On the one hand, in the user's insurance type tendency identification link, the traditional method relies too much on user's active consultation behavior or limited historical data, and lacks effective analysis means for new users without consultation interaction behavior, browsing users without consultation and other groups. This kind of user often leaves no clear demand trace, and the existing system cannot accurately judge their insurance preference, resulting in a large deviation between the recommended results and the actual demand, and poor user experience.

[0004] On the other hand, in the insurance product attribute presentation and comparison aspect, the existing technology mostly adopts the form of text list or simple parameter comparison, and fails to scientifically classify and visualize the product attributes. Insurance product attributes include common features such as price and guarantee period, as well as unique special guarantee features of each insurance type, and the attributes are complex and diverse in dimension. The traditional presentation method makes it difficult for users to intuitively distinguish the differences between products in basic guarantee ability and special guarantee depth, and the comparison efficiency is low, making it difficult to quickly locate the adapted product.

[0005] Therefore, the present application proposes a big data-based insurance product multi-dimensional comparative analysis method and system to address the deficiencies of the prior art. SUMMARY

[0006] In view of the deficiencies of the prior art, the purpose of the present application is to provide a big data-based insurance product multi-dimensional comparative analysis method and system.

[0007] To achieve the above purpose, the present application provides the following technical solutions: The big data-based insurance product multi-dimensional comparative analysis method specifically includes the following steps: Step 1, user insurance type tendency prediction.

[0008] For whether the user produces consultation interaction behavior, the classification strategy is adopted to predict the method, for the user with consultation interaction behavior, the risk matching score is calculated based on the keyword-risk weight matrix; for the user without consultation interaction behavior, the near neighbor user group is matched, and the group preference risk is used as the prediction result; Step two, insurance product visual positioning.

[0009] S21, for the user's tendency risk predicted, the attributes of the insurance products contained in the risk are divided into common attributes and unique attributes; S22, the common attributes and unique attributes of the insurance products are respectively converted into horizontal and vertical coordinates through the index quantification and weighting calculation; S23, the first quadrant is used as the visualization area, and the product position is marked according to the converted horizontal and vertical coordinates; Step three, personalized demand matching area generation.

[0010] The concern degree score of the common attributes and unique attributes in the user demand questionnaire is taken as the core input, the abstract demand score is converted into the personalized demand matching area on the two-dimensional coordinate; Step four, insurance product recommendation set generation.

[0011] The products in the personalized demand matching area are sorted according to the total score of horizontal and vertical coordinates, and written into the insurance product recommendation set, and the maximum characteristics and the maximum shortcomings are marked.

[0012] The application also provides a big data-based insurance product multi-dimensional comparison and analysis system, which comprises a user risk tendency prediction module, an insurance product visual positioning module, a personalized demand matching area generation module and an insurance product recommendation set generation module; The user risk tendency prediction module: the module adopts a differentiated method to predict the risk preference of users with different interaction behaviors; for the user with consultation interaction behavior, the risk matching score is calculated based on the keyword-risk weight matrix; for the user without consultation interaction behavior, the near neighbor user group is matched, and the group preference risk is used as the prediction result; The insurance product visual positioning module: the module directly presents the product positioning in the two-dimensional coordinate area through attribute quantification and coordinate mapping; the product attributes of the user's tendency risk are divided into common attributes and unique attributes; The common attributes are converted into horizontal coordinates, representing the basic guarantee force, and the unique attributes are converted into vertical coordinates, representing the special guarantee depth; the product position is marked in the first quadrant area of the coordinate axis, and the relative performance of the product in the basic guarantee force and the special guarantee depth is directly distinguished; Personalized Needs Matching Area Generation Module: This module constructs a unique matching range based on users' real needs; based on a needs questionnaire survey of users, it obtains the user's concern scores for each attribute, calculates the weight of each attribute based on the concern scores, and determines the minimum acceptable standard for each attribute; based on the weights and the minimum scores, it calculates the minimum values ​​of the horizontal and vertical axes to determine the range of personalized needs matching. Insurance product recommendation collection generation module: This module is the core of the system service output; First, it extracts products within the personalized needs matching area and sorts them from high to low according to the total score of the horizontal and vertical axes; Then, for the sorted products, it marks the biggest feature and the biggest drawback of the product based on the weighted attribute score.

[0013] Furthermore, in step one, the predictive method for directly calculating the insurance type matching score based on the keyword-insurance type weight matrix is ​​as follows: This study focuses on the historical consultation records of different users and the corresponding types of insurance they ultimately purchased; it uses NLP tools to segment the consultation text into words; it groups the segmented consultation texts by insurance type, and counts the top N most frequent words in each group, calculating keywords through mutual information. Insurance types correlation strength ; The association strength is normalized to obtain the weight MI of different insurance types corresponding to each keyword, and finally a keyword-insurance type weight matrix is ​​formed. When a user makes a consultation, the keywords in their consultation text are extracted in real time, the keyword-insurance type weight matrix is ​​queried, the insurance type weights corresponding to the keywords are accumulated, and the matching score of the user for different insurance types is obtained. The insurance type with the highest score is determined as the insurance type that the user prefers.

[0014] Furthermore, in step one, matching nearby user groups and using the group's preferred insurance type as the prediction result, the specific process is as follows: The user did not engage in any inquiry behavior; a user feature vector was constructed based on two types of features collected from big data. ; The cosine similarity formula is used to calculate the feature similarity between the target user and other users; the top-K users with the highest similarity are selected to form a nearest neighbor group; the insurance purchase records of the nearest neighbor group are statistically analyzed to calculate the preference probability of each type of insurance; the insurance type with the highest-1 preference probability is taken as the prediction result of the user's preferred insurance type.

[0015] Furthermore, in step S21, the common attributes include reasonable pricing, flexible coverage period, claims efficiency, and the reputation of the underwriting company, while the unique attributes include the number of critical illnesses covered by health insurance, the reimbursement ratio for minor / major illnesses, the level of green channel service for medical treatment, and the scope of reimbursement for externally purchased drugs.

[0016] Further, in step S22, the attribute is converted into the horizontal and vertical coordinates, and the specific method is as follows: The common attribute is converted into the horizontal coordinate by introducing the basic guarantee force score: the score standard of each common attribute is set, the weight is given according to the importance of the common attribute, the weighted total score is mapped to the horizontal coordinate in proportion, and the mapped horizontal coordinate represents the basic guarantee force; The specific attribute is converted into the vertical coordinate by introducing the special guarantee depth score: the specific attribute of the insurance product in the user's preferred risk type is selected, the score standard of each specific attribute is set, the weight is given according to the importance of the specific attribute, and the weighted total score is mapped to the vertical coordinate in proportion, and the mapped vertical coordinate represents the special guarantee depth.

[0017] Further, the personalized demand matching area is generated, and the specific generation process is as follows: S31, each question in the user demand questionnaire is marked with a corresponding attribute label; the care degree description selected by the user is converted into a care degree score; S32, the single care degree proportion of the attribute is calculated as the attribute weight; and the minimum acceptable score of the attribute is determined; S33, the coordinate value of each common attribute = the attribute score x 20; , Wherein, The number of common attributes is determined, and the horizontal coordinate range of the personalized demand matching area is: ; The coordinate value of each specific attribute = the attribute score x 20; , Wherein, The number of specific attributes is determined, and the vertical coordinate range of the personalized demand matching area is: ; S34, the intersection of the X-axis interval and the Y-axis interval forms a rectangular area as the personalized demand matching area.

[0018] Further, in step four, the maximum feature is specifically the attribute with the highest score x weight selected from the common attribute and the specific attribute, the product introduction of the corresponding insurance product is called from the official website or the background data, and the advantage performance of the attribute is described; and the maximum disadvantage is specifically the attribute with the lowest score x weight selected from the common attribute and the specific attribute, and the short board performance is described by comparing the product introduction of other insurance products.

[0019] Compared with the prior art, the present application has the following beneficial effects: 1. Accurate identification of user's insurance tendency: For users with consultation interaction behavior, the consultation text is analyzed by NLP technology, and the matching score is calculated by combining keyword-insurance weight matrix; for users without consultation behavior, a vector is constructed based on behavior and scene characteristics, and the cosine similarity is matched to the adjacent group preference. This differentiated prediction method breaks through the limitations of traditional single data dependence, and realizes accurate locking of insurance tendency whether the user produces interaction behavior or not, improving the universality and accuracy of demand positioning; 2. Intuitive product multi-dimensional comparison experience: The insurance product attributes are creatively divided into common attributes (price reasonableness, claim efficiency, etc.) and unique attributes (disease coverage of critical illness, vehicle damage insurance coverage, etc.), and are converted into horizontal and vertical coordinates through weighted quantification, with the first quadrant coordinate area marking the product position; intuitively present the relative performance of the product in basic guarantee and special guarantee depth, solve the problem of complex attributes, difficult to distinguish between good and bad in traditional text list or parameter comparison, let users quickly perceive product differences. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 A flowchart of the insurance product multi-dimensional comparison analysis method based on big data; Figure 2 A schematic diagram of the insurance product presented in the coordinate area; Figure 3 A module block diagram of the insurance product multi-dimensional comparison analysis system based on big data. DETAILED DESCRIPTION

[0021] Embodiment one, refer to Figure 1 The insurance product multi-dimensional comparison analysis method based on big data of the embodiment, specifically includes the following steps: Step one, user insurance tendency prediction.

[0022] According to whether the user produces consultation interaction behavior, a classified strategy prediction method is adopted, for users with consultation interaction behavior: the keywords in the consultation text are mined through semantic understanding, and the insurance matching score is directly calculated based on the keyword-insurance weight matrix; for users without consultation interaction behavior: a feature-based collaborative filtering algorithm is adopted, and the group preference insurance is used as the prediction result by matching the adjacent user group through multi-dimensional user characteristics; S11, insurance tendency prediction of users with consultation interaction behavior: Focusing on the historical consultation records of different users (including online customer service dialogues, form consultations, in-app consultation messages, etc.) and the corresponding final insurance types (personal insurance / property insurance / auto insurance, etc.), it is necessary to ensure that each consultation record is clearly labeled with the insurance result tag; among them, each user has one historical consultation record, and the number of historical consultation records needs to be sufficient to ensure the reliability of subsequent statistical methods. In this embodiment, the number of historical consultation records is 3000. Using NLP tools, the consultation text is segmented into words, such as splitting "I want to consult about critical illness insurance and accident insurance" into words like "consult / critical illness insurance / accident insurance". The segmented consultation text is then grouped according to the type of insurance (personal insurance, car insurance, property insurance, etc.), and the top N most frequent words in each group are counted (N is set to 20 in this example to avoid missing information). Examples are shown in Table 1.

[0023] Keyword calculation using mutual information Insurance types correlation strength The calculation formula is as follows: ; in, For words Insurance types The probability of co-occurrence is calculated using the word... Insurance types The number of times they appear together is divided by the total number of historical consultation records. For words The probability of occurrence of a word is calculated using the word... The number of occurrences divided by the total number of historical consultation records; For insurance types The overall selection probability is calculated by selecting the type of insurance. The number of historical consultation records divided by the total number of historical consultation records; Because the mutual information values ​​between different keywords and insurance types may vary in magnitude, normalization is required to ensure that the weights fall within the 0-1 range, thus guaranteeing that higher association strength corresponds to higher weights; for example, "protection" in life insurance... In property insurance In car insurance ,but , Its weighting in life insurance (MI) is Weight in property insurance for This visually reflects the differences in the strength of association; ultimately, a keyword-insurance type weight matrix is ​​formed, serving as the basis for matching; an example of the keyword-insurance type weight matrix is ​​shown in Table 2:

[0024] When a user makes a consultation, keywords are extracted from their consultation text in real time. The keyword-insurance type weight matrix is ​​then queried, and the insurance type weights corresponding to the keywords are summed. The calculation formula is as follows: ; in, This article provides a collection of keywords for user inquiries. For users For insurance types The matching score is used to predict the type of insurance that the user prefers. S12. Prediction of insurance product preferences without consultation interaction: Users may not have made any inquiries, but they may have browsing, registration information, or historical behavior data on the platform (such as newly registered users who have not made inquiries). Based on big data collection, two types of features are used to construct user feature vectors, and the specific dimensions are shown in Table 3:

[0025] Min-Max normalization is applied to numerical features (such as browsing time): ,in, It represents the normalized value of a specific feature within a certain numerical feature. A specific feature value representing a certain numerical characteristic. This represents the minimum value of this type of numerical feature among multiple users. This represents the maximum value of this numerical feature among multiple users. One-hot encoding is used for categorical features (such as occupation): Each category is transformed into a new binary feature. If a user belongs to a certain category, the corresponding feature value is 1; otherwise, it is 0. For example, for occupation categories white-collar, blue-collar, and freelance, when a user's occupation is white-collar, the feature value is 1; the feature values ​​for blue-collar and freelance are both 0. All core features are concatenated to form a user feature vector. ( (The total feature dimension is 6 in this embodiment). Constructing a higher-dimensional user feature vector can yield more accurate matching results. Calculate target users using the cosine similarity formula Other users Feature similarity (of users with known insurance results obtained through big data): ; in, Indicates user User feature vectors, Indicates user User feature vectors, This indicates taking the vector * Norm; The top-K users with the highest similarity (K=50 in this example) are selected to form a nearest neighbor group. Ensure that the distribution of group characteristics is consistent with the target users; statistically analyze neighboring groups. Based on the insurance records, calculate the probability of preference for each type of insurance: ; in, Indicates user For insurance types The probability of preference; The insurance types with the highest-1 preference probability are taken as the predicted insurance types that users prefer; Step 2: Visual positioning of insurance products.

[0026] S21. Classification of Insurance Product Attributes: For the insurance types that users are expected to prefer, the attributes of the insurance products included in the insurance type are divided into common attributes and unique attributes. Common attributes are fundamental characteristics shared by all insurance products, reflecting the product's universal value and basic protection capabilities. These mainly include the following core indicators: Price rationality: The comparison between the product premium and the average price of similar products reflects the economic cost-effectiveness of the premium; Flexibility of coverage period: The range of optional coverage periods, whether renewal or period adjustment is supported, etc., reflect the product's ability to adapt to different needs and cycles; Claims efficiency: The average time from reporting an incident to receiving compensation, and the degree of simplification of the claims process, reflect the convenience of the service; Underwriting company reputation: The insurance company's solvency rating, industry reputation score, complaint rate, etc., reflect the safety and reliability of the product; Unique attributes: Unique attributes are the core characteristics that are unique to a specific type of insurance, reflecting the product's specialized protection value in a specific segment. The unique attributes of different types of insurance vary significantly. Health insurance: Number of critical illnesses covered, reimbursement ratio for minor / major illnesses, level of green channel service for medical treatment, scope of reimbursement for externally purchased drugs, etc.; Car insurance: maximum coverage for vehicle damage insurance, reimbursement ratio for theft insurance, coverage tiers for third-party liability insurance, number of roadside assistance services, deductible for scratch insurance, etc. Property insurance: Coverage of property types, compensation ratio for natural disasters, deductible setting, efficiency of subrogation services, etc. S22, attribute to coordinate conversion: common attributes and unique attributes of insurance products are respectively converted into horizontal and vertical coordinates through index quantification and weighted calculation; S221, introduce basic guarantee force score to convert common attributes into horizontal coordinates: A 5-point grading standard is set for each common attribute (1 point is the lowest, and 5 points are the highest), and the source of the score includes but is not limited to big data; Price reasonableness: take the industry monthly average price of the same type and same amount of product as the benchmark, 5 points for premium below 20% of the average price and above, 4 points for premium below 10%-20% of the average price, 3 points for premium within the average price ± 10% interval, 2 points for premium above 10%-20% of the average price, and 1 point for premium above 20% of the average price. If the product is a combination insurance, calculate separately after splitting the core guarantee responsibility and then take the weighted average value; Guarantee period flexibility: support 5 or more guarantee period options (such as 1 year / 5 years / 10 years / 20 years / lifetime) and support no review of renewal, 5 points; support 3-4 guarantee period options and renewal requires notification but no waiting period, 4 points; support 2 guarantee period options and renewal requires re-underwriting, 3 points; only support one fixed guarantee period and do not support renewal, 2 points; guarantee period cannot be selected, 1 point; Claim efficiency: average claim to account time ≤ 1 working day and support full online claim process, 5 points; average claim to account time 2-3 working days and 80% of the process can be completed online, 4 points; average claim to account time 4-7 working days and some offline materials are required, 3 points; average claim to account time 8-15 working days, 2 points; average claim to account time > 15 working days or multiple materials are required, 1 point; Underwriting company credibility: China Banking and Insurance Regulatory Commission (CBIRC) pays ability rating AAA level and industry complaint rate below average 50%, 5 points; pays ability AA level and complaint rate below average 30%, 4 points; pays ability A level and complaint rate within average ± 20% interval, 3 points; pays ability BB level and complaint rate above average 30%, 2 points; pays ability B level and below or complaint rate above average 50%, 1 point; According to the importance of common attributes, weights can be assigned according to industry knowledge or user-defined, in this embodiment, weights are set according to industry knowledge: price reasonableness weight 0.3, guarantee period flexibility 0.2, claim efficiency 0.3, and underwriting company credibility 0.2; the weighted total score is mapped to the horizontal coordinate in proportion (5-point system corresponds to weighted total score x 20), and the mapped horizontal coordinate represents the basic guarantee force, the higher the value, the better the basic experience of the product; S222, introduce special guarantee depth score to convert unique attributes into vertical coordinates: The specific attributes of the insurance product in the user's preferred risk category are selected by market research, reference to industry standards, expert weight assignment, etc. Referring to step S221, a 5-point scoring standard is set for each common attribute (1 point is the lowest and 5 points is the highest), and the scoring sources include but are not limited to big data; according to the importance of the specific attributes, weights are assigned, which can be determined according to industry knowledge or user-defined; the weighted total score is mapped to the ordinate in proportion (5-point system corresponds to weighted total score x 20), and the mapped ordinate represents the depth of special protection, and the higher the value, the stronger the product's protection capability in the sub-field; Taking health insurance as an example, the depth of special protection calculation process is as follows: a. Select specific attributes: number of critical illness coverage, number of light illness claims, and hospital green pass service level, and user-defined weights (APP or web page interaction) are 0.4, 0.3, and 0.3, respectively; b. Index quantification: Number of critical illness coverage: Count the number of critical illness coverage of all insurance products under health insurance, and sort them from most to least; the top 20% of products get 5 points; the 20%-40% of products get 4 points; the 40%-60% of products get 3 points; the 60%-80% of products get 2 points; the last 20% of products get 1 point; Number of light illness claims: 5 claims without interval get 5 points; 3-4 claims without interval get 4 points; 3 claims or 2 claims without interval get 3 points; 2 claims get 2 points; only 1 claim get 1 point; Hospital green pass service level: includes hospitalization arrangement in Top 50 hospitals nationwide and expert second diagnosis and treatment, gets 5 points; includes hospitalization arrangement in Top 100 hospitals nationwide and expert outpatient service, gets 4 points; includes outpatient appointment in provincial capital city third-grade hospitals and rapid examination, gets 3 points; only provides telephone consultation and ordinary registration assistance, gets 2 points; no green pass service, gets 1 point; c. Weighted calculation: a health insurance product covers 110 critical illnesses (4 points), has 4 light illness claims without grouping (4 points), and has green pass with hospitalization arrangement in Top 100 hospitals nationwide (4 points), so the weighted total score is: 4x0.4+4x0.3+4x0.3=4 points; d. Score conversion: ordinate value = 4x20 = 80 points, representing that the product's depth of special protection in the health insurance field reaches 80 points, at a high level; S23, coordinate area presentation: as shown in Figure 2 , the first quadrant (both horizontal and vertical coordinates are in the range of 0-100) is used as the visualization area, and the product position is directly marked, and the specific presentation details are as follows: Coordinate scale: the horizontal coordinate is evenly distributed from left to right in the range of 0-100, and the basic protection force is marked; the vertical coordinate is evenly distributed from bottom to top in the range of 0-100, and the depth of special protection is marked; Product labeling: Each product is marked with a solid pattern, and the product name is written next to the mark; Reference line settings: Add dashed reference lines at the 50-minute mark on both the horizontal and vertical axes. The intersection of the dashed lines represents the industry average baseline, making it easy to visually assess a product's performance relative to the industry benchmark: Products located to the upper right of the reference line have better basic coverage and deeper specialized coverage than the industry average; products located to the upper left of the reference line have outstanding specialized coverage but need improvement in basic coverage; products located to the lower right of the reference line have solid basic coverage but weak specialized coverage; products located to the lower left of the reference line have both indicators below the industry benchmark. Step 3: Generate personalized demand matching area.

[0027] Users need to complete a needs survey questionnaire to measure their current product needs. The questionnaire includes descriptions of their concern for the common and unique attributes of this type of insurance product, which users can select. Using the user's concern rating for common and unique attributes in the needs survey questionnaire as the core input, the abstract needs rating is transformed into a personalized needs matching area on a two-dimensional coordinate system. S31. Questionnaire data structuring: Standardize the common and unique attributes of the needs survey questionnaire for concern data; assign corresponding attribute labels to each question in the questionnaire to represent the correlation between the two; convert the concern descriptions selected by users (such as very concerned, somewhat concerned, neutral, not very concerned, not concerned) into a 5-point concern score, where 5 points represent very concerned and 1 point represents not concerned. S32. Weight allocation based on level of interest: First, calculate the percentage of interest in each item. The calculation method is as follows: ; Then, determine the minimum acceptable score for the attribute, calculated as follows: ; The higher the level of interest in an attribute, the stricter the minimum standard. The method for calculating the weight of unique attributes is as follows: ; The minimum acceptable score rules share common attributes; S33, Coordinate Range Mapping: Determining the basic guarantee range: The coordinate value of each common attribute = attribute score × 20; , in, The x-axis range of the personalized requirement matching area is: representing the number of common attributes. ; The special guarantee depth interval is determined by using the same calculation logic as the horizontal coordinate, , wherein, the number of unique attributes, the vertical coordinate range of the personalized demand matching area is: ; S34, personalized demand matching area generation and verification: The intersection of the X-axis interval and the Y-axis interval forms a rectangular area: ; If the number of products in the area is less than P (indicating that the number of available products is insufficient, P is 3 in this embodiment), the minimum score of the attribute with the lowest interest degree is reduced (such as from 3 points to 2 points), and the interval is recalculated. If the number of products in the area is greater than Q (indicating that the selection is overloaded, Q is 15 in this embodiment), the minimum score of the attribute with the highest interest degree is increased (such as from 4 points to 5 points), and the area range is reduced. Step four, insurance product recommendation set generation.

[0028] The products in the personalized demand matching area are sorted according to the horizontal and vertical coordinate total score values and written into the insurance product recommendation set, and the maximum feature (the attribute with the highest "score x weight" is selected from the common attributes and unique attributes, the product description of the insurance product is retrieved from the official website or background data, and the advantage performance of the attribute is described) and the maximum disadvantage (the attribute with the lowest "score x weight" is selected from the common attributes and unique attributes, and the shortcoming performance is described by comparing other insurance products).

[0029] Embodiment two, refer to Figure 2 The insurance product multi-dimensional comparison analysis system based on big data in this embodiment includes a user insurance type tendency prediction module, an insurance product visual positioning module, a personalized demand matching area generation module, and an insurance product recommendation set generation module. User insurance type tendency prediction module: This module is the basis for the system to accurately position user demand, and the core function is to use differentiated methods to predict the insurance type preference of users with different interactive behaviors: For users with consultation interactive behavior: Focus on their historical consultation records, use NLP tools to segment the consultation text, count high-frequency words, and use mutual information formula to calculate the correlation strength of keywords and insurance types, and build a keyword-insurance type weight matrix. When the user generates a new consultation, extract the keywords and calculate the weight score based on the matrix, and take the insurance type with the highest score as the tendency prediction result; For users without consulting interaction behavior: Based on user behavior characteristics and scene characteristics to construct feature vector, process features through Min-Max normalization and one-hot encoding; Adopt cosine similarity to calculate the similarity between target user and known insurance record users, select Top-K neighbor users to form a group, and count the insurance preference probability of the group insurance record, and take the highest probability insurance as the inclination prediction result; Insurance product visual positioning module: This module realizes the intuitive presentation of insurance products, and the advantages and disadvantages of products can be intuitively perceived through attribute quantification and coordinate mapping: Insurance product attribute classification: The product attributes of the user's preferred insurance are divided into common attributes and unique attributes; Attribute to coordinate conversion: Common attributes are converted into horizontal coordinates (representing basic protection force), and unique attributes are converted into vertical coordinates (representing special protection depth), and the weighted total score is mapped to the 0-100 interval in proportion; Coordinate region presentation: Label product location in the first quadrant visualization area to intuitively distinguish the relative performance of products in basic protection force and special protection depth; Personalized demand matching area generation module: This module is based on user's real demand to build exclusive matching range, to ensure that the recommended results are highly consistent with user's expectations: Questionnaire data structured processing: Score the concern degree of common attributes and unique attributes in the user demand questionnaire, and label the corresponding attribute for the question; Weight and interval calculation: Calculate the weight of each attribute according to the concern degree score, and determine the minimum acceptable standard of the attribute; Based on the weight and minimum score, calculate the minimum value of the horizontal and vertical coordinates to determine the personalized demand matching interval range; Region generation and verification: Form a rectangular matching area with the intersection of coordinate intervals, and dynamically adjust to ensure the reasonableness of the area; Insurance product recommendation set generation module: This module is the core of the system service output, providing accurate and transparent product recommendations for users; First, extract the products within the personalized demand matching area, and sort them by horizontal and vertical coordinate total score from high to low; Then, for the sorted products, select the attribute with the highest "score x weight" from the common attributes and unique attributes as the biggest feature, and call the product description to explain the advantages; Select the attribute with the lowest "score x weight" as the biggest disadvantage, and compare it with other products to explain the shortcoming.

[0030] Through the detailed introduction of the above-mentioned embodiments, the insurance product multi-dimensional comparison and analysis method and system based on big data can accurately lock the risk tendency of the user, improve the universality and accuracy of demand positioning, and intuitively present the relative performance of the product in the basic guarantee and the depth of the special guarantee.

[0031] The above formulas are all dimensionless values, and the preset parameters in the formulas are set by a person skilled in the art according to actual conditions.

[0032] The above-mentioned embodiments can be realized by software, hardware, firmware or any combination thereof, in whole or in part. When realized by software, the above-mentioned embodiments can be realized in the form of a computer program product in whole or in part. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another, for example, the computer instructions can be transferred from one website, computer, server or data center to another website, computer, server or data center through wired or wireless (such as infrared, wireless, microwave, etc.) mode. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. containing one or more available medium sets. The available medium can be a magnetic medium (such as a floppy disk, a hard disk, a magnetic tape), an optical medium (such as a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state disk.

[0033] It should be understood that in various embodiments of the present application, the size of the sequence number of the above-mentioned processes does not mean the order of execution, and the execution order of the processes should be determined by its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0034] Those skilled in the art can clearly understand that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0035] Those skilled in the art can clearly understand that, for the convenience and brevity of the description, the specific working processes of the above-described system, device and unit can refer to the corresponding processes in the foregoing method embodiments, which will not be repeated here.

[0036] In several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented in other ways. For example, the above-described device embodiments are only schematic, for example, the division of the units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interface, device or unit, and can be electrical, mechanical or other forms.

[0037] If the functions are realized in the form of software function units and sold or used as independent products, they can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the parts that make contributions to the prior art or parts of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0038] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A multi-dimensional comparative analysis method for insurance products based on big data, characterized in that, The method flow is as follows: Step 1: Predicting User Insurance Type Preferences: For users with consultation interaction behavior, calculate the insurance type matching score based on the keyword-insurance type weight matrix; for users without consultation interaction behavior, match with neighboring user groups, and use the group's preferred insurance types as the prediction result. Step Two: Visual Positioning of Insurance Products S21. For the predicted user preference for certain types of insurance, the attributes of the insurance products included in the insurance types are divided into common attributes and unique attributes. S22. The common and unique attributes of insurance products are quantified by indicators and converted into horizontal and vertical axes by weighted calculation. S23. Use the first quadrant as the visualization area and mark the product location based on the horizontal and vertical coordinates of the conversion. Step 3: Generation of Personalized Needs Matching Area: Using the user needs survey questionnaire's rating of concern for common and unique attributes as the core input, the needs rating is transformed into a personalized needs matching area on a two-dimensional coordinate system. Step 4: Generate the insurance product recommendation set: Sort the products within the personalized needs matching area according to the total score of the horizontal and vertical axes and write them into the insurance product recommendation set. Then, mark the biggest feature and the biggest drawback of the product according to the weighted score of each attribute.

2. The method for multi-dimensional comparative analysis of insurance products based on big data according to claim 1, characterized in that, Step one, the calculation of the insurance type matching score based on the keyword-insurance type weight matrix, is as follows: This study focuses on the historical consultation records of different users and the corresponding types of insurance they ultimately purchased; it uses NLP tools to segment the consultation text into words; it groups the segmented consultation texts by insurance type, and counts the top N most frequent words in each group, calculating keywords through mutual information. Insurance types correlation strength ; The association strength is normalized to obtain the weight MI of different insurance types corresponding to each keyword, forming a keyword-insurance type weight matrix; When a user makes a consultation interaction, the keywords in their consultation text are extracted in real time, the keyword-insurance type weight matrix is ​​queried, the insurance type weights corresponding to the keywords are accumulated, and the matching score of the user for different insurance types is obtained. The insurance type with the highest score is determined as the insurance type that the user prefers.

3. The method for multi-dimensional comparative analysis of insurance products based on big data according to claim 1, characterized in that, Step one, matching nearby user groups, uses the group's preferred insurance type as the prediction result. The specific process is as follows: When users do not engage in consultation or interaction, user feature vectors are constructed based on two types of features collected from big data. The cosine similarity formula is used to calculate the feature similarity between the target user and other users; the top-K users with the highest similarity are selected to form a nearest neighbor group; the insurance purchase records of the nearest neighbor group are statistically analyzed to calculate the preference probability of each type of insurance; the insurance type with the highest-1 preference probability is taken as the prediction result of the user's preferred insurance type.

4. The method for multi-dimensional comparative analysis of insurance products based on big data according to claim 1, characterized in that, In step S21, the common attributes include reasonable pricing, flexible coverage period, claims efficiency, and the reputation of the underwriting company. The specific attributes include the number of critical illnesses covered by health insurance, the reimbursement ratio for minor / major illnesses, the level of green channel medical services, and the scope of reimbursement for externally purchased drugs.

5. The method for multi-dimensional comparative analysis of insurance products based on big data according to claim 1, characterized in that, In step S22, the specific method for converting the attributes into horizontal and vertical coordinates is as follows: The basic protection capability score is introduced to transform common attributes into a horizontal axis: a scoring standard is set for each common attribute, weights are assigned according to the importance of the common attributes, and the weighted total score is mapped to the horizontal axis proportionally. The mapped horizontal axis represents the basic protection capability. The introduction of a specialized protection depth score transforms unique attributes into a vertical axis: Selecting unique attributes of insurance products among the types of insurance that users prefer, setting scoring criteria for each unique attribute, assigning weights according to the importance of the unique attributes, and mapping the weighted total score proportionally to the vertical axis, with the mapped vertical axis representing the specialized protection depth.

6. The method for multi-dimensional comparative analysis of insurance products based on big data according to claim 1, characterized in that, The personalized needs matching area is generated through the following process: S31. Label each question in the user needs survey questionnaire with its corresponding attribute tag; convert the user's selected level of concern into a level of concern score; S32. Calculate the percentage of individual interest in an attribute, and use it as the attribute weight; Determine the minimum acceptable score for each attribute; S33, The coordinate values ​​of each common attribute are mapped to... interval; ， in, The x-axis range of the personalized needs matching area is determined by the number of common attributes: ; The coordinate values ​​of each unique attribute are mapped to... interval; ， in, The vertical axis range, representing the number of unique attributes, determines the area for matching personalized needs: ; S34. The rectangular area formed by the intersection of the X-axis interval and the Y-axis interval is used as the personalized demand matching area.

7. The method for multi-dimensional comparative analysis of insurance products based on big data according to claim 1, characterized in that, In step four, the most significant feature is to select the attribute with the highest weighted score from the common and unique attributes, retrieve the product introduction of the corresponding insurance product from the official website or backend data, and describe the advantages of the attribute; the most significant disadvantage is to select the attribute with the lowest weighted score from the common and unique attributes, compare it with the product introduction of other insurance products, and describe its shortcomings.

8. A big data-based multidimensional comparative analysis system for insurance products, used to implement the big data-based multidimensional comparative analysis method for insurance products as described in any one of claims 1-7, characterized in that, The system includes a user insurance type preference prediction module, an insurance product visualization and positioning module, a personalized demand matching area generation module, and an insurance product recommendation set generation module. User Insurance Preference Prediction Module: This module uses differentiated methods to predict users' insurance preferences based on different interactive behaviors; for users with consultation interaction behaviors, it calculates insurance matching scores based on keyword-insurance weight matrix; For users who do not engage in consultation or interaction, they are matched with nearby user groups, and the group's preferred insurance types are used as the prediction result. Insurance Product Visualization and Positioning Module: This module presents the product positioning intuitively in a two-dimensional coordinate area through attribute quantification and coordinate mapping. First, the product attributes of the insurance types that users prefer are divided into common attributes and unique attributes. Then, the common attributes are transformed into the horizontal axis, representing the basic protection level, and the unique attributes are transformed into the vertical axis, representing the depth of specific protection. Finally, the product position is marked in the first quadrant of the coordinate axis, intuitively distinguishing the relative performance of the product in terms of basic protection level and depth of specific protection. Personalized Needs Matching Area Generation Module: This module constructs a unique matching range based on users' real needs; based on a needs questionnaire survey of users, it obtains the user's concern scores for each attribute, calculates the weight of each attribute based on the concern scores, and determines the minimum acceptable standard for each attribute; based on the weights and the minimum scores, it calculates the minimum values ​​of the horizontal and vertical axes to determine the range of personalized needs matching. Insurance product recommendation collection generation module: This module is the core of the system service output; First, it extracts products within the personalized needs matching area and sorts them from high to low according to the total score of the horizontal and vertical axes; Then, for the sorted products, it marks the biggest feature and the biggest drawback of the product based on the weighted attribute score.

Images