Insurance matching method based on artificial intelligence

Through Gaussian regularization constraints and sparse subspace clustering optimization, combined with neural network and clustering layer, using thermonuclear similarity and Laplace regularization terms, the impact of existing insurance matching methods in noise and outliers and local effects and outliers are solved. The problem in structural capture improves the accuracy and speed of insurance matching.

CN118674562BActive Publication Date: 2024-11-15SHENZHEN YINSHUO TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411155571.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-22
Publication Date
2024-11-15
Estimated Expiration
2044-08-22

AI Technical Summary

Technical Problem

The existing insurance matching methods are susceptible to noise and outliers, and cannot determine the potential structure of the data, poor model interpretation ability, resulting in poor insurance matching effect, and inability to capture local structural information, poor adaptability of loss function, and slow parameter update speed.

Method used

Introduce Gaussian regularization constraints and sparse subspace clustering optimization, build models based on neural network layer and clustering layer, calculate similarity using thermocore similarity, introduce Laplace regularization terms and adaptive parameters, and design iterative update strategies .

Benefits of technology

It improves the robustness of the model to noise and outliers, enhances data smoothness and clustering effect, captures local structural information, and improves the accuracy and speed of insurance matching.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118674562B_ABST
    Figure CN118674562B_ABST
Patent Text Reader

Abstract

The present invention discloses an insurance matching method based on artificial intelligence, and the method includes data collection, data preprocessing, data optimization, establishing an insurance matching model and insurance matching. The present invention belongs to the field of insurance matching technology, and specifically refers to an insurance matching method based on artificial intelligence. This scheme improves the quality of insurance matching data by introducing Gaussian regularization constraints and sparse subspace clustering optimization; thereby improving the accuracy of insurance matching; constructing an insurance matching model based on a neural network layer and a clustering layer, using hot kernel similarity to calculate the similarity between insurance matching data, and improving the local accuracy of clustering; introducing a Laplace regularization term to enhance the robustness and accuracy of the model, and adaptive parameters to ensure the flexibility of the loss function when processing the distribution of different insurance matching data; weighting factors are used to adjust the influence of the distance between the sample and the cluster center, thereby making the insurance matching accurate and fast.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of insurance matching, and in particular to an insurance matching method based on artificial intelligence. Background Art

[0002] The insurance matching method uses big data technology to optimize the matching process between insurance products and customer needs; this method aims to improve the accuracy and efficiency of insurance product recommendations, improve customer experience, and achieve personalized services. However, the general insurance matching method is easily affected by noise and outliers in the insurance matching data, unable to determine the underlying structure of the data, and the model has poor interpretation ability for insurance matching data, which leads to poor insurance matching results; the general insurance matching method is unable to capture the local structural information between insurance matching sample data, the model's loss function has poor adaptability to insurance matching data, and the model parameter update speed is slow, which leads to poor final insurance matching results. Summary of the invention

[0003] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides an insurance matching method based on artificial intelligence. In view of the problem that general insurance matching methods are easily affected by noise and outliers in insurance matching data, cannot determine the potential structure of the data, and the model has poor interpretation ability of insurance matching data, which leads to poor insurance matching effect, this scheme improves the quality of insurance matching data by introducing Gaussian regularization constraints and sparse subspace clustering optimization. Gaussian regularization constraints can make the model more robust to noise and outliers in insurance matching data, reduce the impact of abnormal data on the model; enhance data smoothness and improve clustering effect; by optimizing the processing of noise in the data through sparse subspace clustering, the efficiency of insurance matching data processing and the interpretation ability of the model are improved, so that the core structure of the insurance matching data is more prominent; thereby improving insurance Accuracy of matching; In view of the problems that general insurance matching methods cannot capture the local structural information between insurance matching sample data, the model's loss function has poor adaptability to insurance matching data, and the model parameter update speed is slow, which leads to poor final insurance matching effect, this scheme builds an insurance matching model based on neural network layer and clustering layer, and uses hot kernel similarity to calculate the similarity between insurance matching data, which helps to capture the local structural information between insurance matching data and improve the local accuracy of clustering; the introduction of Laplace regularization term enhances the robustness and accuracy of the model, and the adaptive parameters ensure the flexibility of the loss function in dealing with the distribution of different insurance matching data; the weighting factor is used to adjust the influence of the distance between the sample and the cluster center to improve the accuracy of the model; the model convergence speed is accelerated by designing an iterative update strategy, thereby making the insurance matching accurate and fast.

[0004] The technical solution adopted by the present invention is as follows: The insurance matching method based on artificial intelligence provided by the present invention comprises the following steps:

[0005] Step S1: data collection;

[0006] Step S2: data preprocessing;

[0007] Step S3: data optimization;

[0008] Step S4: Establishing an insurance matching model;

[0009] Step S5: Insurance matching.

[0010] Furthermore, in step S1, the data collection is to collect historical insurance matching data and real-time insurance matching data; the historical insurance matching data and the real-time insurance matching data both include customer information, insurance needs, historical claim records and customer preferences; the historical insurance matching data also includes customer satisfaction; customer satisfaction is used as a data label; data labels are only used for cluster label selection and do not participate in calculations as data dimensions, and the label with the largest number of historical insurance matching data is selected as the cluster label; the cluster label to which the real-time insurance matching data belongs is used as the prediction result of the real-time insurance matching data.

[0011] Furthermore, in step S2, the data preprocessing is to clean, convert and standardize the collected data; the data cleaning is to process missing values, duplicate values ​​and outliers; the data conversion is to convert the cleaned data into a vector form; and the standardization is based on the maximum and minimum normalization method.

[0012] Furthermore, in step S3, the data optimization specifically includes the following steps:

[0013] Step S31: introduce Gaussian regularization constraints; the form of Gaussian regularization constraints is as follows:

[0014] ;

[0015] ;

[0016] In the formula, is a given mean set M and covariance matrix , the Gaussian distribution probability density function value of the random variable Z; Z is a set of n1 random variables; z1, z2 and z n1 are the first, second and n1th random variables respectively; M is the vector of N1 means A set of components, each mean vector corresponds to the mean of a Gaussian distribution; is the inverse matrix of the covariance matrix; T is the transpose operation;

[0017] Step S32: Optimize the sparse subspace clustering to obtain the objective function, obtain the sparse representation of the data, and obtain the optimized data set; the objective function is as follows:

[0018] ;

[0019] ;

[0020] In the formula, is the L1 norm; is the square of the Frobenius norm; and is the regularization parameter; and is the Gaussian regularization constraint matrix; is a non-negative regularization parameter; is the element-by-element multiplication; A is the auxiliary matrix; diag(·) is the diagonalization operation on the matrix; E is the error matrix; X is the data matrix.

[0021] Furthermore, in step S4, the establishment of the insurance matching model is to process the historical insurance matching data optimized in step S3, and to divide the historical insurance matching data into a test set and a training set in advance; specifically, the following steps are included:

[0022] Step S41: Initialization; the insurance matching model includes a neural network layer and a clustering layer; the neural network layer is used to receive the historical insurance matching data and the real-time insurance matching data after data optimization, and the data is converted into a feature representation suitable for clustering through the encoder in the neural network; the membership matrix U is initialized so that the membership of each sample satisfies the constraint condition ; For each layer of the neural network, randomly initialize the weight matrix W(l) and bias b(l) of the encoder and decoder; where u nc is the membership degree of the nth sample to the cth cluster; C is the total number of clusters;

[0023] Step S42: define the objective function; specifically including:

[0024] Step S421: Calculate the heat kernel similarity; expressed as follows:

[0025] ;

[0026] In the formula, s ij is the similarity between the i-th sample and the j-th sample; x i and x j are the i-th sample and the j-th sample respectively; NB i is the neighborhood of the i-th sample;

[0027] Step S422: constructing a clustering loss function;

[0028] Step S423: define the objective function of the insurance matching model; it is expressed as follows:

[0029] ;

[0030] ;

[0031] In the formula, and X n are the predicted sample data and the actual sample data respectively; γ is the regularization parameter; L is the number of neural network layers; l is the layer index; is the weighting factor; is the Frobenius norm squared;

[0032] Step S43: Update network parameters; based on the objective function of the insurance matching model, use the Adam optimization algorithm to optimize the network parameters;

[0033] Step S44: updating the cluster center matrix; specifically including:

[0034] Step S441: define the cluster center objective function; it is expressed as follows:

[0035] ;

[0036] In the formula, is the cluster center objective function;

[0037] Step S442: Gradient calculation; expressed as follows:

[0038] ;

[0039] Step S443: Update the formula; set the gradient to zero, and obtain:

[0040] ;

[0041] In the formula, is the updated cluster center;

[0042] Step S45: updating the membership matrix; specifically including:

[0043] Step S451: Define membership objective function ; means as follows:

[0044] ;

[0045] In the formula, is the Lagrange multiplier;

[0046] Step S452: Gradient calculation; expressed as follows:

[0047] ;

[0048] Step S453: Solve; get:

[0049] ;

[0050] ;

[0051] ;

[0052] In the formula, is the updated membership degree; and It is an auxiliary item;

[0053] Step S454: Update the Lagrange multiplier; expressed as follows:

[0054] ;

[0055] In the formula, is the updated Lagrange multiplier;

[0056] Step S455: Substitute the updated membership degree; it is expressed as follows:

[0057] ;

[0058] Step S46: judgment; when the insurance matching model converges to the objective function of the training set, the insurance matching model training is completed. A threshold is set in advance. If the prediction accuracy of the insurance matching model for the test set is higher than the threshold, the insurance matching model is established; if the maximum number of training times is reached, the data set is re-divided to train the model; otherwise, return to step S43 to continue iteration.

[0059] Furthermore, in step S5, the insurance matching is based on the established insurance matching model and real-time insurance matching data, and a customer satisfaction threshold is set in advance. If the customer satisfaction predicted by the insurance matching model for the real-time insurance matching data is higher than the customer satisfaction threshold, the insurance type corresponding to the real-time insurance matching data is matched to the customer.

[0060] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0061] (1) In view of the problem that general insurance matching methods are easily affected by noise and outliers in insurance matching data, cannot determine the underlying structure of the data, and have poor model interpretation of insurance matching data, which leads to poor insurance matching results, this scheme improves the quality of insurance matching data by introducing Gaussian regularization constraints and sparse subspace clustering optimization. Gaussian regularization constraints can make the model more robust to noise and outliers in insurance matching data, reduce the impact of abnormal data on the model; enhance data smoothness and improve clustering effects; and improve the efficiency of insurance matching data processing and the model's interpretation ability by optimizing sparse subspace clustering to process noise in data, making the core structure of insurance matching data more prominent; thereby improving the accuracy of insurance matching.

[0062] (2) In view of the problems that general insurance matching methods cannot capture the local structural information between insurance matching sample data, the model's loss function has poor adaptability to insurance matching data, and the model parameter update speed is slow, which leads to poor final insurance matching effect. This scheme constructs an insurance matching model based on a neural network layer and a clustering layer, and uses hot kernel similarity to calculate the similarity between insurance matching data, which helps to capture the local structural information between insurance matching data and improve the local accuracy of clustering; the introduction of Laplace regularization terms enhances the robustness and accuracy of the model, and the adaptive parameters ensure the flexibility of the loss function in dealing with the distribution of different insurance matching data; the weighting factor is used to adjust the influence of the distance between the sample and the cluster center to improve the accuracy of the model; the model convergence speed is accelerated by designing an iterative update strategy, thereby making the insurance matching accurate and fast. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 A schematic diagram of the process of the insurance matching method based on artificial intelligence provided by the present invention;

[0064] Figure 2 is a schematic diagram of the process of step S3;

[0065] Figure 3 It is a schematic diagram of the process of step S4.

[0066] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0067] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0068] In the description of the present invention, it should be understood that terms such as “upper”, “lower”, “front”, “back”, “left”, “right”, “top”, “bottom”, “inside” and “outside” indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as limiting the present invention.

[0069] Example 1, see Figure 1 The insurance matching method based on artificial intelligence provided by the present invention comprises the following steps:

[0070] Step S1: data collection, collecting historical insurance matching data and real-time insurance matching data;

[0071] Step S2: Data preprocessing, data cleaning, data conversion and standardization of the collected data;

[0072] Step S3: data optimization, which is achieved by introducing Gaussian regularization constraints and sparse subspace clustering optimization;

[0073] Step S4: Establish an insurance matching model. The insurance matching model is constructed based on the neural network layer and the clustering layer. The similarity between the insurance matching data is calculated using the heat kernel similarity. The Laplace regularization term and the adaptive parameter are introduced to design the loss function. The insurance matching model is established based on the weighting factor and the design iterative update strategy.

[0074] Step S5: Insurance matching.

[0075] Example 2, see Figure 1 This embodiment is based on the above embodiment. In step S1, both the historical insurance matching data and the real-time insurance matching data include customer information, insurance demand, historical claim records and customer preferences; the historical insurance matching data also includes customer satisfaction; customer satisfaction is used as a data label; the data label is only used for cluster label selection and does not participate in the calculation as a data dimension. The label with the largest number of historical insurance matching data is selected as the cluster label; the cluster label to which the real-time insurance matching data belongs is used as the prediction result of the real-time insurance matching data.

[0076] Example 3, see Figure 1This embodiment is based on the above embodiment. In step S2, data cleaning is to process missing values, duplicate values ​​and outliers; data conversion is to convert the cleaned data into a vector form; and standardization is based on the maximum and minimum normalization method.

[0077] Example 4, see Figure 1 and Figure 2 This embodiment is based on the above embodiment. In step S3, data optimization specifically includes the following steps:

[0078] Step S31: introduce Gaussian regularization constraints; the form of Gaussian regularization constraints is as follows:

[0079] ;

[0080] ;

[0081] In the formula, is a given mean set M and covariance matrix , the Gaussian distribution probability density function value of the random variable Z; Z is a set of n1 random variables; z1, z2 and z n1 are the first, second and n1th random variables respectively; M is the vector of N1 means A set of components, each mean vector corresponds to the mean of a Gaussian distribution; is the inverse matrix of the covariance matrix; T is the transpose operation;

[0082] Step S32: Optimize the sparse subspace clustering to obtain the objective function, obtain the sparse representation of the data, and obtain the optimized data set; the objective function is as follows:

[0083] ;

[0084] ;

[0085] In the formula, is the L1 norm; is the square of the Frobenius norm; and is the regularization parameter; and is the Gaussian regularization constraint matrix; is a non-negative regularization parameter; is the element-by-element multiplication; A is the auxiliary matrix; diag(·) is the diagonalization operation on the matrix; E is the error matrix; X is the data matrix.

[0086] By performing the above operations, the general insurance matching method is easily affected by the noise and outliers in the insurance matching data, unable to determine the potential structure of the data, and the model has poor explanatory ability for the insurance matching data, which leads to poor insurance matching effect. This scheme improves the quality of insurance matching data by introducing Gaussian regularization constraints and sparse subspace clustering optimization. Gaussian regularization constraints can make the model more robust to noise and outliers in the insurance matching data, reduce the impact of abnormal data on the model; enhance data smoothness and improve clustering effect; by optimizing the sparse subspace clustering to process the noise in the data, the efficiency of insurance matching data processing and the explanatory ability of the model are improved, making the core structure of the insurance matching data more prominent; thereby improving the accuracy of insurance matching.

[0087] Example 5, see Figure 1 and Figure 3 This embodiment is based on the above embodiment. In step S4, the insurance matching model is established by processing the historical insurance matching data optimized in step S3, and the historical insurance matching data is pre-divided into a test set and a training set; specifically, the following steps are included:

[0088] Step S41: Initialization; the insurance matching model includes a neural network layer and a clustering layer; the neural network layer is used to receive the historical insurance matching data and the real-time insurance matching data after data optimization, and the data is converted into a feature representation suitable for clustering through the encoder in the neural network; the membership matrix U is initialized so that the membership of each sample satisfies the constraint condition ; For each layer of the neural network, randomly initialize the weight matrix W(l) and bias b(l) of the encoder and decoder; where u nc is the membership degree of the nth sample to the cth cluster; C is the total number of clusters;

[0089] Step S42: define the objective function; specifically including:

[0090] Step S421: Calculate the heat kernel similarity; expressed as follows:

[0091] ;

[0092] In the formula, s ij is the similarity between the i-th sample and the j-th sample; x i and x j are the i-th sample and the j-th sample respectively; NB i is the neighborhood of the i-th sample;

[0093] Step S422: constructing a clustering loss function;

[0094] Step S423: define the objective function of the insurance matching model; it is expressed as follows:

[0095] ;

[0096] ;

[0097] In the formula, and X n are the predicted sample data and the actual sample data respectively; γ is the regularization parameter; L is the number of neural network layers; l is the layer index; is the weighting factor; is the Frobenius norm squared;

[0098] Step S43: Update network parameters; based on the objective function of the insurance matching model, use the Adam optimization algorithm to optimize the network parameters;

[0099] Step S44: updating the cluster center matrix; specifically including:

[0100] Step S441: define the cluster center objective function; it is expressed as follows:

[0101] ;

[0102] In the formula, is the cluster center objective function;

[0103] Step S442: Gradient calculation; expressed as follows:

[0104] ;

[0105] Step S443: Update the formula; set the gradient to zero, and obtain:

[0106] ;

[0107] In the formula, is the updated cluster center;

[0108] Step S45: updating the membership matrix; specifically including:

[0109] Step S451: define the membership objective function ; means as follows:

[0110] ;

[0111] In the formula, is the Lagrange multiplier;

[0112] Step S452: Gradient calculation; expressed as follows:

[0113] ;

[0114] Step S453: Solve; get:

[0115] ;

[0116] ;

[0117] ;

[0118] In the formula, is the updated membership degree; and It is an auxiliary item;

[0119] Step S454: Update the Lagrange multiplier; expressed as follows:

[0120] ;

[0121] In the formula, is the updated Lagrange multiplier;

[0122] Step S455: Substitute the updated membership degree; it is expressed as follows:

[0123] ;

[0124] Step S46: judgment; when the insurance matching model converges to the objective function of the training set, the insurance matching model training is completed. A threshold is set in advance. If the prediction accuracy of the insurance matching model for the test set is higher than the threshold, the insurance matching model is established; if the maximum number of training times is reached, the data set is re-divided to train the model; otherwise, return to step S43 to continue iteration.

[0125] By performing the above operations, the general insurance matching method cannot capture the local structural information between the insurance matching sample data, the model's loss function has poor adaptability to the insurance matching data, and the model parameter update speed is slow, which leads to the final poor insurance matching effect. This scheme builds an insurance matching model based on the neural network layer and the clustering layer, and uses the hot kernel similarity to calculate the similarity between the insurance matching data, which helps to capture the local structural information between the insurance matching data and improve the local accuracy of clustering; the introduction of the Laplace regularization term enhances the robustness and accuracy of the model, and the adaptive parameters ensure the flexibility of the loss function in dealing with the distribution of different insurance matching data; the weighting factor is used to adjust the influence of the distance between the sample and the cluster center to improve the accuracy of the model; the model convergence speed is accelerated by designing an iterative update strategy, thereby making the insurance matching accurate and fast.

[0126] Example 6, see Figure 1This embodiment is based on the above embodiment. In step S5, insurance matching is based on the established insurance matching model and real-time insurance matching data, and a customer satisfaction threshold is pre-set. If the customer satisfaction predicted by the insurance matching model for the real-time insurance matching data is higher than the customer satisfaction threshold, the insurance type corresponding to the real-time insurance matching data is matched to the customer.

[0127] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0128] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the invention.

[0129] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. Insurance matching method based on artificial intelligence, Features: The method comprises the following steps: Step S1: data collection; Step S2: data preprocessing; Step S3: data optimization, which is achieved by introducing Gaussian regularization constraints and sparse subspace clustering optimization; Step S4: Establish an insurance matching model. The insurance matching model is constructed based on the neural network layer and the clustering layer. The similarity between the insurance matching data is calculated using the heat kernel similarity. The Laplace regularization term and the adaptive parameter are introduced to design the loss function. The insurance matching model is established based on the weighting factor and the design iterative update strategy. Step S5: insurance matching; Step S4 includes step S42: defining the objective function; step S42 includes step S422: constructing a clustering loss function; introducing Laplace regularization; expressed as follows: ; ; In the formula, is the clustering loss function; V is the cluster center matrix; N is the total number of samples; x n is the nth sample; v c is the cth cluster center; τ is an adaptive parameter used to measure the distance from the sample to the cluster center, ensuring that the loss function can be adjusted according to the distance between the sample and the cluster center; is the weight parameter of the Laplace regularization term; is the neighborhood of the nth sample; s nj is the similarity between the nth sample and the jth sample; u jc is the membership of the jth sample to the cth cluster; U is the membership matrix; C is the total number of clusters; u nc is the membership degree of the nth sample to the cth cluster.

2. The insurance matching method based on artificial intelligence according to claim 1, Features: In step S3, the data optimization specifically includes the following steps: Step S31: introduce Gaussian regularization constraints; the form of Gaussian regularization constraints is as follows: ; ; In the formula, is a given mean set M and covariance matrix , the Gaussian distribution probability density function value of the random variable Z; Z is a set of n1 random variables; z 1 、z 2 and z n1 are the first, second and n1th random variables respectively; M is the vector of N1 means A set of components, each mean vector corresponds to the mean of a Gaussian distribution; is the inverse of the covariance matrix; T is the transpose operation; Step S32: Optimize the sparse subspace clustering to obtain the objective function, obtain the sparse representation of the data, and obtain the optimized data set; the objective function is as follows: ; ; In the formula, is the L1 norm; is the square of the Frobenius norm; and is the regularization parameter; and is the Gaussian regularization constraint matrix; is a non-negative regularization parameter; is the element-by-element multiplication; A is the auxiliary matrix; diag(·) is the diagonalization operation on the matrix; E is the error matrix; X is the data matrix.

3. The insurance matching method based on artificial intelligence according to claim 1, Features: In step S4, the establishment of the insurance matching model is to process the historical insurance matching data optimized in step S3, and to divide the historical insurance matching data into a test set and a training set in advance; specifically, the following steps are included: Step S41: Initialization; the insurance matching model includes a neural network layer and a clustering layer; the neural network layer is used to receive the historical insurance matching data and the real-time insurance matching data after data optimization, and the data is converted into a feature representation suitable for clustering through the encoder in the neural network; the membership matrix U is initialized so that the membership of each sample satisfies the constraint condition ; For each layer of the neural network, randomly initialize the weight matrix W(l) and bias b(l) of the encoder and decoder; where u nc is the membership degree of the nth sample to the cth cluster; C is the total number of clusters; Step S42: define the objective function; specifically including: Step S421: Calculate the heat kernel similarity; expressed as follows: ; In the formula, s ij is the similarity between the i-th sample and the j-th sample; x i and x j are the i-th sample and the j-th sample respectively; NB i is the neighborhood of the i-th sample; Step S422: constructing a clustering loss function; Step S423: define the objective function of the insurance matching model; it is expressed as follows: ; ; In the formula, and X n are the predicted sample data and the actual sample data respectively; γ is the regularization parameter; L is the number of neural network layers; l is the layer index; is the weighting factor; is the Frobenius norm squared; Step S43: Update network parameters; based on the objective function of the insurance matching model, use the Adam optimization algorithm to optimize the network parameters; Step S44: updating the cluster center matrix; specifically including: Step S441: define the cluster center objective function; it is expressed as follows: ; In the formula, is the cluster center objective function; Step S442: Gradient calculation; expressed as follows: ; Step S443: Update the formula; set the gradient to zero, and obtain: ; In the formula, is the updated cluster center; Step S45: updating the membership matrix; Step S46: judgment; when the insurance matching model converges to the objective function of the training set, the insurance matching model training is completed. A threshold is set in advance. If the prediction accuracy of the insurance matching model for the test set is higher than the threshold, the insurance matching model is established; if the maximum number of training times is reached, the data set is re-divided to train the model; otherwise, return to step S43 to continue iteration.

4. The insurance matching method based on artificial intelligence according to claim 3, Features: In step S45, updating the membership matrix specifically includes: Step S451: define the membership objective function ; means as follows: ; In the formula, is the Lagrange multiplier; Step S452: Gradient calculation; expressed as follows: ; Step S453: Solve; get: ; ; ; In the formula, is the updated membership degree; and It is an auxiliary item; Step S454: Update the Lagrange multiplier; expressed as follows: ; In the formula, is the updated Lagrange multiplier; Step S455: Substitute the updated membership degree; expressed as follows: 。 5. The insurance matching method based on artificial intelligence according to claim 1, Features: In step S1, the data collection is to collect historical insurance matching data and real-time insurance matching data; the historical insurance matching data and the real-time insurance matching data both include customer information, insurance needs, historical claim records and customer preferences; the historical insurance matching data also includes customer satisfaction; customer satisfaction is used as a data label; data labels are only used for cluster label selection and do not participate in calculations as data dimensions, and the labels with the largest number of historical insurance matching data are selected as cluster labels; the cluster labels to which real-time insurance matching data belongs are used as prediction results for real-time insurance matching data.

6. The insurance matching method based on artificial intelligence according to claim 1, Features: In step S5, the insurance matching is based on the established insurance matching model and real-time insurance matching data, and a customer satisfaction threshold is set in advance. If the customer satisfaction predicted by the insurance matching model for the real-time insurance matching data is higher than the customer satisfaction threshold, the insurance type corresponding to the real-time insurance matching data is matched to the customer.

7. The insurance matching method based on artificial intelligence according to claim 1, Features: In step S2, the data preprocessing is to clean, convert and standardize the collected data; the data cleaning is to process missing values, duplicate values ​​and abnormal values; The data conversion is to convert the cleaned data into a vector form; the standardization processing is based on the maximum and minimum normalization method to perform standardization processing.