Existing house insurance underwriting risk estimation method based on support vector machine

Abstract

The invention discloses an existing house insurance underwriting risk estimation method based on a support vector machine, and relates to the technical field of house physical examination and insurance risk analysis, and the method comprises the steps: carrying out the house structure detection and evaluation of an existing house project, and obtaining a house physical examination report; analyzing the relative importance weight of each risk factor of the house insurance item according to the house physical examination report; calculating a house insurance underwriting risk level; inputting the relative importance weight of each risk factor of the house insurance item of the to-be-detected house into the trained classification model based on the vector machine to obtain a house insurance underwriting risk level of the to-be-detected house; according to the method, quantitative processing is performed in combination with various risk factors of house physical examination and house insurance items and the house insurance underwriting risk level, and finally, the house insurance underwriting risk level estimation is performed through the classification model trained based on the vector machine, so that the influence of subjective factors is reduced, and the estimation result is more accurate.

Description

Technical Field

[0001] The present invention relates to the technical field of housing physical examination and insurance risk analysis, and in particular to a method for predicting the underwriting risk of existing housing insurance based on a support vector machine. Background Art

[0002] With the development of the urbanization process, a large number of existing buildings have been built for more than a certain number of years (such as 30 years) and have entered the middle and late stages of their service life. Many old buildings have the characteristics of "low construction standards, lack of necessary maintenance, and huge potential safety hazards". As time goes by, the problems will become increasingly serious. It is proposed to establish three systems of housing physical examination, insurance, and pension, and carry out insurance-related pilot projects on housing safety issues.

[0003] Currently, the research on housing insurance focuses on the insurance for potential defects in the quality of newly built houses, and there is less research on the safety insurance of existing houses. Moreover, in the field of engineering insurance risk prediction, currently, the method of directly evaluating the indicators in the risk assessment system by experts is mostly used, and the risk analysis results are often greatly affected by the subjective factors of experts. Summary of the Invention

[0004] In order to overcome the deficiencies in the existing technology that there is less research on the underwriting risk prediction of existing housing insurance and the evaluation accuracy is low, the present invention proposes a method for predicting the underwriting risk of existing housing insurance based on a support vector machine.

[0005] To achieve the above object, the present invention adopts the following technical solutions, including:

[0006] S1. Conduct housing structure detection and evaluation on the existing housing project to obtain a housing physical examination report;

[0007] S2: Analyze the relative importance weights of each risk factor of the housing insurance project according to the housing physical examination report;

[0008] S3: Calculate the housing insurance underwriting risk level;

[0009] S4: Based on the relative importance weights of each risk factor of the housing insurance project and the housing insurance underwriting risk level, construct a housing insurance underwriting risk data set;

[0010] S5: Construct a basic classification model, use the relative importance weights of each risk factor of the housing insurance project as the input of the model, and the housing insurance underwriting risk level as the output of the model, and train the model to obtain a trained classification model;

[0011] S6: Input the relative importance weights of each risk factor of the housing insurance project of the housing to be tested into the trained classification model to obtain the housing insurance underwriting risk level of the housing to be tested.

[0012] Preferably, in step S2, analyze the relative importance weights of each risk factor of the housing insurance project according to the housing inspection report, including:

[0013] S21: Obtain the housing design data, and set the risk factors of the housing insurance project according to the housing design data and the housing inspection report;

[0014] S22: For each risk factor, use the risk index matrix evaluation method to determine the risk factor level;

[0015] S23: Based on the levels of each risk factor of the housing insurance project, use the analytic hierarchy process to determine the relative importance weights of each risk factor.

[0016] Preferably, in step S23, based on the levels of each risk factor of the housing insurance project, use the analytic hierarchy process to determine the relative importance weights of each risk factor, including:

[0017] S231: Based on the levels of each risk factor of the housing insurance project, construct a judgment matrix Q = (q ij ) m×m ;

[0018] S232: Based on the judgment matrix, construct a column-normalized matrix

[0019] S233: Obtain the relative importance weights w i of each risk factor based on the column-normalized matrix, and the calculation formula is as follows:

[0020]

[0021] where i and j are the i-th and j-th risk factors respectively, q ij , are the element values of the i-th row and j-th column in the judgment matrix Q and the column-normalized matrix respectively, and m is the total number of risk factors.

[0022] Preferably, after obtaining the relative importance weights of each risk factor, first obtain the maximum eigenvalue λ max of the judgment matrix Q, and the calculation formula is:

[0023]

[0024] Then, based on the maximum eigenvalue of the judgment matrix Q, obtain the consistency index CI = (λ max - m) / (m - 1);

[0025] Then calculate the random consistency ratio CR = CI / RI, where RI is the random consistency index;

[0026] Judge whether the random consistency ratio CR is less than 0.1; if so, the calculation results of the relative importance weights of each risk factor are reasonable; if not, adjust the judgment matrix Q until the condition is met.

[0027] Preferably, the element value q ij in the judgment matrix Q is calculated by the following formula:

[0028] q ij = 2(x i - x j ) + 1

[0029] where x i and x j are the levels of the i-th and j-th risk factors respectively.

[0030] Preferably, the construction method of the column normalization matrix is as follows:

[0031]

[0032] Preferably, in step S3, calculating the housing insurance underwriting risk level includes:

[0033] S31: Obtain the claim amount and insurance amount of the housing insurance project;

[0034] S32: Based on the claim amount and insurance amount of the housing insurance project, obtain the housing insurance underwriting risk level F, and the calculation formula is:

[0035]

[0036] where a is the ratio of the claim amount to the insurance amount of the housing insurance project, and a1, a2, a3, a4 are set thresholds that increase in sequence.

[0037] Preferably, the basic classification model is a classification model based on a support vector machine, and the calculation method of the loss function S of the trained classification model is:

[0038]

[0039] The constraint condition is: Y k (r·X k + b) ≥ 1;

[0040] where X k is the relative importance weight of each risk factor of the k-th sample housing insurance project, f(.) is the classification function, and the output is the predicted value of the housing insurance underwriting risk level, Y k is the housing insurance underwriting risk level of the k-th sample, N is the total number of samples, C is the penalty factor, is the regularization term, It is the hinge loss, where r and b are the normal vector and intercept of the hyperplane respectively.

[0041] Preferably, the training of the basic classification model is based on the Radial Basis Function (RBF) to obtain the trained classification model. The expression of the radial basis function K(X, X′) is:

[0042] K(X, X′) = exp(-γ·||X - X′|| 2 )

[0043] where exp(.) represents the exponential function with the natural constant e as the base, X and X′ represent any two input variables of the basic classification model, that is, the relative importance weights of the risk factors of two sample housing insurance items, γ represents the kernel function coefficient; ||X - X′|| 2 represents the Euclidean distance between the two input variables. The radial basis function K(X, X′) maps the risk factor importance weights to a high-dimensional space by calculating the Euclidean distance between the two input variables and exponentiating, realizing non-linear classification.

[0044] The advantages of the present invention are as follows:

[0045] (1) By conducting housing structure detection and evaluation on existing houses, obtaining a housing physical examination report, analyzing the relative importance weights of the risk factors of housing insurance items according to the housing physical examination report, and then using the trained classification model based on the support vector machine, the housing insurance underwriting risk level of the house to be measured is obtained. Combining the housing physical examination, the risk factors of housing insurance items, and the housing insurance underwriting risk level for quantification processing, and finally predicting the housing insurance underwriting risk level through the trained classification model based on the support vector machine. This method can unify the underwriting risk analysis process of each housing insurance item, reduce the influence of subjective factors, and make the prediction result more accurate.

[0046] (2) By conducting a housing physical examination on existing houses, evaluating the risk factors of housing insurance from the housing physical examination report, and using the Analytic Hierarchy Process (AHP) to analyze the relative importance weights of the risk factors, and using the existing housing insurance cases as the training data set to train the classification model for predicting the insurance risk level, and outputting the predicted existing housing insurance underwriting risk level, so as to realize the scientific formulation and optimization of housing insurance, adapt to the market environment and different housing conditions, and have strong scientificity and flexibility.

[0047] (3) The risk factors of housing insurance items are set according to the housing design data and the housing physical examination report. For each risk factor, the risk factor matrix evaluation method is used to determine the risk factor level, making the risk factor level scientific and reasonable, thus promoting the more reasonable and accurate prediction of the existing housing insurance underwriting risk.

[0048] (4) Based on the risk factor levels of the housing insurance project, the present invention constructs a judgment matrix, constructs a column-normalized matrix based on the judgment matrix, obtains the relative importance weights of each risk factor based on the column-normalized matrix, and then combines with a classification model based on a support vector machine to obtain the underwriting risk level of existing housing insurance, which is convenient to operate and has a low time complexity. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a flowchart of a method for predicting the underwriting risk of existing housing insurance based on a support vector machine according to the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0050] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a 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 those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0051] As Figure 1 shown, the present invention proposes a method for predicting the underwriting risk of existing housing insurance based on a support vector machine, including:

[0052] S1, conduct housing structure detection and evaluation on the existing housing project to obtain a housing physical examination report;

[0053] The housing structure detection includes regular physical examination and emergency physical examination. The regular physical examination refers to the regular physical examination of a house in normal use, and the emergency physical examination refers to the immediate physical examination of a house affected by natural disasters, human damage and other factors. In this embodiment, it is mainly the regular physical examination;

[0054] The housing structure detection of the housing project is completed by a building detection agency, which is jointly confirmed by the owner, the insurance company and the community street;

[0055] The housing structure detection consists of three levels: component and connection detection, subsystem detection, and building system detection. Component and connection detection includes the quality and performance of raw materials of the house, the strength of component materials, component connection nodes, and whether the components are deformed and damaged; the subsystem includes the structural layout and system of the main structure of the house, displacement and deformation, structural disaster prevention performance, and survey data of the foundation and foundation, foundation bearing capacity, deformation and settlement conditions; among them, the damaged parts of the structure modified by humans should be marked separately; the detection results of the building system are determined by the detection results of the components and connections and the subsystems.

[0056] When conducting a housing structure inspection on a housing project, the building inspection agency shall delimit the safety level of the housing structure. In this embodiment, reference can be made to the safety levels of components, sub-units, and appraisal units in the "Standard for Appraisal of Reliability of Civil Buildings" (GB50292-2015), which are A(a)u, B(b)u, C(c)u, and D(d)u, or the building inspection agency can delimit the safety level of the housing structure by itself and make a detailed description of the delimitation basis in the housing physical examination report.

[0057] After the housing structure inspection and evaluation of the housing project are completed, the building inspection agency shall issue a housing physical examination report, which shall be confirmed and passed by the owner, the insurance company, and the community street.

[0058] S2: Analyze the relative importance weights of various risk factors of the housing insurance project according to the housing physical examination report;

[0059] S21: Obtain the housing design data, and set the risk factors of the housing insurance project according to the housing design data and the housing physical examination report; the risk factors of the housing insurance project include initial defect risk, housing environment risk, natural disaster and accident risk, building damage risk, and human transformation risk.

[0060] The initial defect risk refers to the risk that due to the limitations of the housing construction age conditions, improper exploration and site selection, structural design defects, and insufficient construction technology may affect the housing construction quality and safety.

[0061] The housing environment risk refers to the risk that the housing construction quality and safety are affected by the long-term action of environmental factors such as geology, hydrology, topography and geomorphology, and climate in the geographical location where the housing is located.

[0062] The natural disaster and accident risk refers to the risk that the housing construction quality and safety are affected by natural disasters such as earthquakes, wind disasters (typhoons, tornadoes, etc.), heavy rains, lightning, and accidental accidents such as fires and explosions, which are determined by the geographical location where the housing is located.

[0063] The building damage risk refers to the risk that the building structure damage caused by various factors during the building use period affects the housing construction quality and safety.

[0064] The human transformation risk refers to the safety risk caused by residents' transformation of the housing, such as demolishing and changing load-bearing walls, adding floors on the roof, and changing building functions, which weakens the original resistance and service performance of the structure.

[0065] S22: For each risk factor, use the risk index matrix evaluation method (RAC) to determine the risk factor level;

[0066] Quantify each risk factor for different housing insurance projects to construct a risk factor matrix, which consists of risk probability, loss, and risk factor level;

[0067] For a certain risk factor, the risk probability and resulting loss are different in different housing insurance project cases, and the corresponding risk factor levels are also different. Therefore, in the present invention, each risk factor for different housing insurance projects is quantified to construct a risk factor matrix, which consists of risk probability, loss, and risk factor level. The construction of the risk factor matrix for different risk factors should consult the opinions of relevant experts. The relevant experts first determine the risk probability and loss of each risk factor for different housing insurance projects, and then evaluate and determine the risk factor level of each risk factor based on the risk probability and loss of each risk factor. The value range of the risk factor level of each risk factor is an integer between 1 and 5;

[0068] The risk factor matrix table is shown in Table 1;

[0069] Table 1 Risk Factor Matrix Table

[0070]

[0071] Table 2 Risk Factor Level Description Table

[0072] Risk factor level Description of risk factor level 1 Negligible risk 2 Tolerable risk 3 Risk that requires attention 4 Risk that requires high attention 5 Unacceptable risk

[0073] S23: Based on the risk factor levels of the housing insurance project, use the analytic hierarchy process to determine the relative importance weights of each risk factor, including:

[0074] S231: Based on the risk factor levels of the housing insurance project, construct a judgment matrix Q=(q ij ) 5×5 ;

[0075]

[0076] Among them, the element value q ij in the judgment matrix Q is calculated by the formula:

[0077] q ij =2(x i -x j )+1

[0078] Among them, x i and x j are the risk factor levels of the i-th and j-th risk factors respectively; the meaning of different values of the element value q ij in the i-th row and j-th column of the judgment matrix Q is shown in Table 3.

[0079] Table 3 Matrix Q Element Value Meaning Table

[0080] <![CDATA[q ij value]]> Meaning 1 Indicates that the importance of two risk factors is the same 3 Indicates that one risk factor is slightly more important than another risk factor 5 Indicates that one risk factor is more important than another risk factor 7 Indicates that one risk factor is significantly more important than another risk factor 9 Indicates that one risk factor is strongly more important than another risk factor 2、4、6、8 Indicates that the importance level is between two adjacent rating scales

[0081] S232: Based on the judgment matrix Q, construct a column-normalized matrix The construction method is as follows:

[0082]

[0083] S233: Obtain the relative importance weight w of each risk factor based on the column-normalized matrix i , and the calculation formula is as follows:

[0084]

[0085] Among them, i and j respectively represent the i-th and j-th risk factors, and q ij , are the element values in the i-th row and j-th column of the judgment matrix Q and the column-normalized matrix respectively.

[0086] After obtaining the relative importance weights of each risk factor, first obtain the maximum eigenvalue λ of the judgment matrix Q max , and the calculation formula is:

[0087]

[0088] Then, based on the maximum eigenvalue of the judgment matrix Q, obtain the consistency index CI = (λ max -5) / (5 - 1); then calculate the random consistency ratio CR = CI / RI, where RI = 1.12;

[0089] Judge whether CR is less than 0.1. If yes, the calculation results of the relative importance weights of each risk factor are reasonable. If not, adjust the judgment matrix Q until the condition is met.

[0090] S3: Calculate the housing insurance underwriting risk level;

[0091] S31: Obtain the claim amount and insurance amount of the housing insurance project;

[0092] S32: Based on the claim amount and insurance amount of the housing insurance project, obtain the housing insurance underwriting risk level F, and the calculation formula is:

[0093]

[0094] Among them, a is the ratio of the claim amount to the insurance amount of the housing insurance project.

[0095] S4: Construct a housing insurance underwriting risk data set based on the relative importance weights of various risk factors of the housing insurance project and the housing insurance underwriting risk level;

[0096] The housing insurance underwriting risk data set includes a training data set and a test data set, and the ratio of the sample numbers is 8:2.

[0097] S5: Construct a basic classification model, use the relative importance weights of various risk factors of the housing insurance project as the input of the model, and the housing insurance underwriting risk level as the output of the model, and train the model to obtain a trained classification model;

[0098] The basic classification model is a classification model based on a support vector machine. It mainly finds an optimal decision boundary (hyperplane) during the data binary classification process, and the boundary is the farthest from the nearest samples of the two categories, so as to maximize the spacing and achieve the best classification effect.

[0099] There are 5 levels of the housing insurance underwriting risk level in the embodiment of the present invention. Using the basic classification model based on the vector machine, the multi-classification problem is transformed into a problem of 5 binary classification variables. Therefore, the output variable is constructed as a 5-dimensional vector Y = {y1, y2,..., y i ,..., y5}, and the value of each dimension can only take 1 or -1. When the value of a dimension is 1, the other dimensions can only take -1. The dimension index with a value of 1 represents the housing insurance underwriting risk level of the sample. For example: Y = {-1, -1, -1, 1, -1} of a sample means that the housing insurance underwriting risk level of the sample is level 4.

[0100] The calculation method of the loss function S of the trained classification model is:

[0101]

[0102] The constraint condition is: Y k (r·X k +b)≥1;

[0103] Among them, X k is the relative importance weights of various risk factors of the kth sample housing insurance project, f(.) is the classification function, and the output is the predicted value of the housing insurance underwriting risk level. Y k is the housing insurance underwriting risk level of the kth sample, N is the total number of samples, C is the penalty factor, is the regularization term, is the hinge loss, and r and b are the normal vector and intercept of the hyperplane respectively.

[0104] The optimization goal of the basic classification model is to maximize the classification spacing and ensure that the model has a certain classification accuracy.

[0105] Select the radial basis kernel function (RBF) to train the basic classification model, and obtain the trained classification model. The expression of the radial basis kernel function K(X, X′) is as follows:

[0106] K(X, X′) = exp(-γ·||X - X′|| 2 )

[0107] where exp(.) represents the exponential function with the natural constant e as the base, X and X′ represent any two input variables of the basic classification model, that is, the relative importance weights of the risk factors of two sample housing insurance items, γ represents the kernel function coefficient; ||X - X′|| 2 represents the Euclidean distance between the two input variables.

[0108] The RBF kernel function measures the similarity between two input variables X and X′ by calculating the Euclidean distance ||X - X′|| 2 . The smaller the distance, the closer the function value is to 1, indicating that the samples are more similar; the larger the distance, the closer the function value is to 0, indicating that the samples are more different. This characteristic enables the RBF kernel function to effectively capture the local similarity between data points. The parameter γ in the function controls the width (i.e., "influence range") of the kernel function. A larger γ value will narrow the response range of the kernel function, and the model will pay more attention to local samples; a smaller γ value will widen the response range, and the model will consider more global features; the RBF kernel function maps the original data to a high-dimensional feature space through an exponential transformation, thereby transforming the originally linearly inseparable problem into a linearly separable problem. This mapping ability makes it particularly suitable for dealing with complex non-linear classification tasks.

[0109] The radial basis kernel function K(X, X′) maps the importance weights of risk factors to a high-dimensional space by calculating the Euclidean distance between two input variables and exponentiating, to achieve non-linear classification.

[0110] S6: Input the relative importance weights of the risk factors of the housing insurance item of the house to be tested into the trained classification model based on the support vector machine, to obtain the housing insurance underwriting risk level of the house to be tested.

[0111] Certainly, for those skilled in the art, the present invention is not limited to the details of the above exemplary embodiments, but also includes the same or similar structures that can be implemented in other specific forms without departing from the spirit or basic characteristics of the present invention. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be included in the present invention. Any reference signs in the claims should not be regarded as limiting the claims involved.

[0112] In addition, it should be understood that although this specification is described according to embodiments, not every embodiment only contains an independent technical solution. This narrative manner of the specification is only for clarity. Those skilled in the art should regard the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

[0113] The technologies, shapes, and structures not detailedly described in the present invention are all well-known technologies.

Claims

1. A method for estimating the underwriting risk of existing houses based on support vector machines, characterized in that Including: S1. Conduct housing structure inspection and assessment on the existing housing project to obtain a housing physical examination report; S2: Analyze the relative importance weights of various risk factors of the housing insurance project according to the housing physical examination report; S3: Calculate the housing insurance underwriting risk level; S4: Based on the relative importance weights of various risk factors of the housing insurance project and the housing insurance underwriting risk level, construct a housing insurance underwriting risk data set; S5: Construct a basic classification model, use the relative importance weights of various risk factors of the housing insurance project as the input of the model, and the housing insurance underwriting risk level as the output of the model, and train the model to obtain a trained classification model; S6: Input the relative importance weights of various risk factors of the housing insurance project of the house to be measured into the trained classification model to obtain the housing insurance underwriting risk level of the house to be measured.

2. The method for predicting the underwriting risk of existing houses insurance based on support vector machine according to claim 1, characterized in that, In step S2, when analyzing the relative importance weights of various risk factors of the housing insurance project according to the housing physical examination report, it includes: S21: Obtain housing design data, and set the risk factors of the housing insurance project according to the housing design data and the housing physical examination report; S22: For each risk factor, use the risk index matrix evaluation method to determine the risk factor level; S23: Based on the levels of various risk factors of the housing insurance project, use the analytic hierarchy process to determine the relative importance weights of various risk factors.

3. The method for predicting the underwriting risk of existing houses insurance based on support vector machine according to claim 2, wherein, In step S23, when using the analytic hierarchy process to determine the relative importance weights of various risk factors based on the levels of various risk factors of the housing insurance project, it includes: S231: Based on the risk factor levels of the housing insurance project, construct a judgment matrix Q = (q ij ) m×m ; S232: Construct a column-normalized matrix based on the judgment matrix S233: Obtain the relative importance weight w of each risk factor based on the column normalization matrix i , and the calculation formula is as follows: where, i and j are the i-th and j-th risk factors respectively, and q ij and are the element values in the i-th row and j-th column of the judgment matrix Q and the column-normalized matrix respectively, and m is the total number of risk factors.

4. The method for predicting the underwriting risk of existing houses insurance based on support vector machine according to claim 3, characterized in that, After obtaining the relative importance weights of various risk factors, first obtain the maximum eigenvalue λ of the judgment matrix Q max , and the calculation formula is: Then, based on the maximum eigenvalue of the judgment matrix Q, the consistency index CI = (λ max - m) / (m - 1); Then calculate the random consistency ratio CR = CI / RI, where RI is the random consistency index; Judge whether the random consistency ratio CR is less than 0.1; if yes, the calculation results of the relative importance weights of various risk factors are reasonable; if not, adjust the judgment matrix Q until the condition is met.

5. The method for estimating the underwriting risk of existing houses insurance based on support vector machine according to claim 3, characterized in that, The element value q in the judgment matrix Q ij is calculated by the following formula: q ij = 2(x i - x j ) + 1 where x i and x j are the risk factor levels of the i-th and j-th items respectively.

6. The method for estimating the underwriting risk of existing houses insurance based on support vector machine according to claim 3, characterized in that, The column normalization matrix is constructed as follows:

7. The method for predicting the underwriting risk of existing houses insurance based on support vector machine according to claim 1, characterized in that, In step S3, when calculating the housing insurance underwriting risk level, it includes: S31: Obtain the claim amount and insurance amount of the housing insurance project; S32: Based on the claim amount and insurance amount of the housing insurance project, obtain the housing insurance underwriting risk level F, and the calculation formula is: where a is the ratio of the claim amount to the insurance amount of the housing insurance project, and a1, a2, a3, a4 are set thresholds that increase in sequence.

8. The method for predicting the underwriting risk of existing houses insurance based on support vector machine according to claim 1, characterized in that, The basic classification model is a classification model based on a support vector machine, and the calculation method of the loss function S of the trained classification model is: The constraint is: Y k (r·X k +b)≥1; Among them, X k is the relative importance weight of each risk factor of the k-th sample housing insurance project, f(.) is the classification function, and the output is the predicted value of the housing insurance underwriting risk level, Y k is the housing insurance underwriting risk level of the k-th sample, N is the total number of samples, C is the penalty factor, is the regularization term, is the hinge loss, and r and b are the normal vector and intercept of the hyperplane respectively.

9. The method for predicting the underwriting risk of existing houses insurance based on support vector machine according to claim 8, characterized in that Select to use the radial basis kernel function (RBF) to train the basic classification model to obtain a trained classification model, and the expression of the radial basis kernel function K(X, X′) is: K(X,X′) = exp(-γ·||X - X′|| 2 ) Among them, exp(.) represents the exponential function with the natural constant e as the base. X and X′ represent any two input variables of the basic classification model, that is, the relative importance weight values of the risk factors of two sample house insurance items. γ represents the kernel function coefficient; ||X - X′|| 2 represents the Euclidean distance between the two input variables. The radial kernel function K(X, X′) maps the risk factor importance weight values to a high-dimensional space by calculating the Euclidean distance between the two input variables and exponentiating, so as to achieve non-linear classification.

Images