Driving behavior insurance rate grade determination method, system and device

Abstract

In order to improve the accuracy of driving behavior premium rate grade judgment, the invention provides a vehicle driving behavior premium rate grade judgment method, and establishes a vehicle premium rate grade judgment method based on a factor analysis algorithm based on driving behavior risk factors of Internet of Vehicles big data. The method comprises the following steps: respectively calculating a driving behavior risk factor value, constructing a public factor based on factor analysis, combining an accumulated variance contribution rate, calculating a comprehensive evaluation value of the public factor, and carrying out driving behavior premium rate grade judgment on a vehicle by utilizing collected vehicle driving related information. Compared with an SVM algorithm, an XGBoost algorithm and a traditional Logistic regression model, an empirical result shows that by means of the judgment method, the accuracy of the judgment result is improved, meanwhile, a vehicle driver can improve the driving behavior in time according to division of risk categories, and safe and healthy life and economic and stable development of people are facilitated.

Description

Technical Field

[0001] The present invention relates to the technical field of Internet of Things, and in particular to a method, system and device for determining a driving behavior insurance premium rate level. Background Art

[0002] Some vehicle drivers have unsafe driving behaviors, which are likely to cause traffic accidents. Different drivers have different driving habits. Determining insurance premiums based on the driver's driving behavior is the future development direction of auto insurance. In the prior art, there are many studies on risk prediction based on driving behavior, and there are also a few studies on driving behavior insurance pricing models. However, at present, there is no effective method for comprehensively evaluating the driver's behavior and comprehensively evaluating and grading the auto insurance premium. When calculating the insurance premium, only considering the vehicle factor and ignoring the driver factor is not conducive to mobilizing the driver's subjective initiative and cultivating the driver's safety awareness, and the premium burden is also unreasonable. Summary of the invention

[0003] In order to overcome the above-mentioned defects of the existing auto insurance model, the purpose of the present invention is to provide a method, system and device for determining the insurance premium level of vehicle driving behavior, thereby improving the rationality and accuracy of the existing auto insurance pricing model.

[0004] On the one hand, the present application provides a method for determining a driving behavior insurance premium rate level, comprising the following steps:

[0005] S01: Collect driving behavior data;

[0006] S02: Calculating a driving behavior risk factor value based on the collected driving behavior data;

[0007] S03: Standardizing the risk factor value;

[0008] S04: Calculate the correlation coefficient matrix A of the risk factor values;

[0009] S05: Determine the number of common factors according to the risk factors after the standardization process;

[0010] S06: Perform orthogonal rotation transformation on the original factor loading matrix;

[0011] S07: Calculate the inverse matrix of the correlation coefficient matrix A;

[0012] S08: Calculate factor scores according to the inverse matrix of the correlation coefficient matrix A, the original factor loading matrix and the risk factor matrix;

[0013] S09: Comprehensively evaluate the common factors according to the factor scores.

[0014] Furthermore, step S01 includes:

[0015] Use the vehicle OBD device data to obtain the vehicle's sudden acceleration information, sudden braking information, sharp turn information, and mileage information;

[0016] Using DSM equipment data, driver status information is obtained, including physical fatigue, number of phone calls, distracted driving, number of smoking, driver abnormal information, and number of times seat belts were not fastened;

[0017] Utilize ADAS equipment data to obtain ADAS assisted driving warning information, including the number of times the turn signal is not used in lane changes or turns as specified, the number of lane departure warnings, the number of frequent lane change warnings, the number of close vehicle distances and forward collision warnings, and the number of times the vehicle fails to pass according to traffic light instructions.

[0018] Further, step S02 includes:

[0019] When calculating the driving behavior risk factor value, the risk factors affecting driving behavior also include driver information, traffic violation data, and accident data, all of which need to be calculated.

[0020] Further, step S03 includes:

[0021] The risk factor values ​​are standardized. Assuming that m drivers are evaluated, the driving behavior of each driver includes q behavior indicators, and the constructed matrix Y is:

[0022] where y i =(y 1i ,y 2i ,...,y mi )', i = 1, 2, ..., q; it needs to be standardized, and the driving behavior risk factor value is standardized using the following formula:

[0023]

[0024] in

[0025] Further, step S04 includes:

[0026] Calculate the correlation coefficient matrix A of risk factors: Calculate the covariance matrix of q behavioral indicator risk factors, assuming that the correlation coefficient matrix is ​​A = (a ij ) indicates that the calculation is as shown in formula (2):

[0027]

[0028] in,

[0029] Among them, d ij It refers to the standardized transformation of the elements of matrix Y, that is, the standardized processing of the driver's driving behavior indicators; according to the formula, a ij =1, and a ij =a ji .

[0030] Further, step S05 includes:

[0031] Determine the number of common factors: Calculate the variance contribution rate and cumulative variance contribution rate of the principal component factor. If the cumulative variance contribution rate reaches 85% or above, then this factor is the principal component factor; Assume that the standardized q-item behavioral indicator risk factor is represented by Y=(y1,y2,…,y q ), independent common factors are represented by F=(F1,F2,…F n ) means, assuming F1, F2, … F n The mean of is 0 and the variance is 1, so the factor model is expressed by formula (3):

[0032] Y=B×F(3)

[0033] in Where B is called the factor loading matrix;

[0034] For q driving behavior risk factors y i (i=1,2,…q), assuming

[0035] Represents the sum of squares of the elements in row i, also a factor y i Since the driving behavior risk factors and common factors have been standardized, and the special factors and common factors are also independent of each other, The larger the value of , the more information contained in the driving behavior risk factor is summarized by n common factors;

[0036] Assumptions That is, the sum of the squares of the elements in the jth column of the factor loading matrix, expressed as g j represents the sum of the variance contributions of common factors to driving behavior risk factors, g j The larger the value, the more important the common factor j is. Formula (6) represents the variance contribution rate of the common factor:

[0037]

[0038] The variance of the risk factor and Where v(y i ) refers to the variance of driving behavior risk factors; The larger the value, the greater the common factor Fj The higher the importance.

[0039] Further, step S06 includes:

[0040] Original factor loading matrix transformation: In order to obtain more accurate analysis results, it is necessary to transform the original factor loading matrix and calculate the factor loading matrix using the principal component estimation method. The m groups of observations with q behavioral indicators are represented by the matrix Y:

[0041] Determine n common factors to describe driving behavior; assume that the characteristic matrix W,

[0042] W satisfies:

[0043] W'W=WW'=I q (7)

[0044] Among them I q is the q-order unit matrix, and W is an orthogonal matrix. Therefore, according to formula (8):

[0045] F'=W'Y(8)

[0046] F'=(f1,f2,...,f q )' is q principal component risk factors; it includes the first n principal component risk factors and qn risk factors. The first n principal component risk factors are selected as common factors, so the W matrix is ​​expressed as:

[0047] W=(w1,w2,...,w n ,w n+1 ,....,w q )=[W (1) ,W (2) ],

[0048]

[0049] therefore,

[0050] F'=(f1,f2,...,f n ,f n+1 ,...,f q )=[F (1) ,F (2) ],So,

[0051] F (1) =(f1,f2,...,f n ),F (2) =(f n+1 ,f n+2 ,...,fq )

[0052] According to formula (7) and formula (8),

[0053]

[0054] According to formula (9), assuming δ = W (2) F (2) =(δ1,δ2,...,δ q )',So,

[0055] Y=W (1) F (1) +δ(10), according to the principal component analysis algorithm, the variance of the first n principal component risk factors is assumed to be:

[0056] λ1,λ2,...,λ n , then assume that:

[0057] According to formula (10), the calculation formula of factor loading matrix B is as follows (11):

[0058] Therefore, the determination of common factors is based on the characteristic root being greater than or equal to 1, and at the same time, the cumulative variance contribution rate of the first n common factors is not less than 85%;

[0059] The original factor loading matrix is ​​B = (b ij ), through orthogonal rotation transformation T, we get the load matrix A = (a ij ),So,

[0060] A=BT, so that the sum of variances of the factor loadings of the common factors reaches the maximum value. The following formula (12) is the sum of variances:

[0061] For the factor loading matrix after orthogonal rotation, The larger it is, the more accurately the information about the top n common risk factors can describe the original information.

[0062] Further, step S08 includes:

[0063] The factor score formula is shown in formula (13):

[0064] Among them, B' is the transpose of the factor loading matrix, A -1 is the inverse matrix of the risk factor correlation coefficient matrix, and Y is the definition in step S03; the scores of the common factors are calculated according to formula (13).

[0065] Further, step S09 includes:

[0066] The comprehensive evaluation is performed by normalizing the weights of the common factors to obtain the comprehensive evaluation value of the common factors; the calculation is shown in formula (14):

[0067] in, Among them, λ i is the variance of the principal component risk factors.

[0068] On the other hand, the present application provides a device for implementing the above-mentioned method for determining the insurance premium rate level of driving behavior, the device comprising:

[0069] Use the vehicle's OBD device to obtain the vehicle's sudden acceleration information, sudden braking information, sharp turn information, and mileage information;

[0070] Use DSM equipment to obtain driver status information, including physical fatigue, number of phone calls, distracted driving, number of smoking, driver abnormal information, and number of times seat belts were not fastened;

[0071] Utilize ADAS equipment to obtain ADAS assisted driving warning information, including the number of times the turn signal is not used in lane change or turning according to regulations, the number of lane deviation warnings, the number of frequent lane change warnings, the number of close vehicle distance and forward collision warnings, and the number of times the traffic light is not used to pass.

[0072] On the other hand, the present application also provides a system for implementing the above-mentioned method for determining the insurance premium rate level of driving behavior, the system comprising:

[0073] The data acquisition module uses different acquisition devices to obtain the original data of vehicle driving behavior;

[0074] Data storage module, which stores the collected data on the cloud platform;

[0075] Data processing module, which performs denoising and standardization on raw data;

[0076] Data calculation module, which calculates the standardized data;

[0077] The result display module displays the final evaluation results in different display formats to the corresponding devices through different display programs. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] Figure 1 A schematic diagram of a process for determining a driving behavior insurance premium rate level according to the present invention;

[0079] Figure 2 This is a schematic diagram of some risk factors that affect driving behavior in a method for determining a driving behavior insurance premium rate level according to the present invention. DETAILED DESCRIPTION

[0080] The following will clearly and completely describe the technical solutions in the embodiments of the present invention 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, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0081] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0082] like Figure 1 The figure is a flowchart of a method for determining the insurance premium rate level of driving behavior, and the specific steps are as follows:

[0083] S01: Using the vehicle OBD device data, obtain the vehicle's sudden acceleration information (including the road and time period), sudden braking information (including the road and time period), sharp turn information (including the road and time period), and mileage information;

[0084] Using DSM equipment data, driver status information is obtained, including physical fatigue (yawning, closing eyes), number of phone calls, distracted driving (number of times looking around, number of times looking down), number of times smoking, driver abnormal information (number of times the driver leaves the driving line of sight), and number of times the seat belt is not fastened;

[0085] Using ADAS equipment data, obtain ADAS assisted driving warning information, including the number of times the turn signal is not used in lane change and turning according to regulations, the number of lane deviation warnings, the number of frequent lane change warnings, the number of close distance and forward collision warnings, and the number of times the traffic lights are not used as instructed (red and yellow lights);

[0086] S02: Calculate the driving behavior risk factor value. Calculate the driving behavior risk factor based on the data collected by the above three devices. Other risk factors that affect driving behavior include driver information, traffic violation data, accident data (such as Figure 2 ) need to be calculated.

[0087] Specifically, three months of data from 2,123 vehicles of a taxi management company were collected for the experiment. According to the above steps S01 and S02, relevant driving behavior risk factors were calculated, totaling 34 indicators.

[0088] S03: Risk factor standardization. According to the risk factor values ​​calculated in the above steps, the driving behavior of 2123 vehicles needs to be evaluated. Each vehicle is represented by a unique ID. The driving behavior of each vehicle includes 34 behavior indicators. The constructed matrix Y is:

[0089] where yi =(y 1i ,y 2i ,...,y mi )', i = 1, 2, ..., q. Where q = 34, m = 2123, it needs to be standardized, and the driving behavior risk factor value is standardized using the following formula, where i = 1, 2, ..., 2123, j = 1, 2, ..., 34.

[0090]

[0091] in

[0092] S04: Calculate the correlation coefficient matrix A of the risk factors, calculate the covariance matrix of the 2123 driving behavior risk factors, and assume that the correlation coefficient matrix is ​​A=(a ij ) indicates that the calculation is as shown in formula (2):

[0093]

[0094] in, Among them, d ij It refers to the standardized transformation of the elements of matrix Y, that is, the standardized processing of the driver's driving behavior indicators; therefore, according to the formula, it can be concluded that a ij =1, and a ij =a ji , where i=1,2,...,2123,j=1,2,...,34.

[0095] S05: Determine the number of common factors, calculate the variance contribution rate and cumulative variance contribution rate of the main factor, if the cumulative variance contribution rate reaches 85% or above, then determine this factor as the main component factor. Assume that the 2123 driving behavior risk factors after standardization are represented by Y=(y1,y2,…,y q ), independent common factors are represented by F=(F1,F2,…F n ) means, assuming F1, F2, … F n The mean of is 0 and the variance is 1, so the factor model is expressed by formula (3).

[0096] Y=B×F(3)

[0097] in Where B is called the factor loading matrix.

[0098] For q driving behavior risk factors y i (i=1,2,…q), where q=2123, assuming

[0099] Indicates that the sum of the squares of the elements in row i is also a factor of y i Since the driving behavior risk factors and common factors have been standardized, and the special factors and common factors are also independent of each other, The larger the value of , the more information contained in the driving behavior risk factor can be summarized by n common factors. Then the risk factor y i 96.79% of the information in can be included by n common factors, where i = 1, 2, ..., q.

[0100] Assumptions That is, the sum of the squares of the elements in the jth column of the factor loading matrix, expressed as g j represents the sum of the variance contributions of common factors to driving behavior risk factors, g j The larger the value, the more important the common factor j is. Therefore, formula (6) represents the variance contribution rate of the common factor.

[0101]

[0102] The variance of the risk factor and Where v(y i ) refers to the variance of driving behavior risk factors, The larger the value, the greater the common factor F j The higher the importance.

[0103] S06: Transformation of the original factor loading matrix. In order to obtain more accurate analysis results, the original loading matrix needs to be transformed. There are many estimation methods for the original factor loading matrix, and the calculation methods are mostly complicated. In the present invention, the principal component estimation method is used to calculate the factor loading matrix. For m groups of observations with q risk factors, the matrix Y is used to represent:

[0104] Determine n common factors to describe driving behavior. Assume that the feature matrix W,

[0105] W satisfies:

[0106] W'W=WW'=I q (7)

[0107] Among them I q is the q-order unit matrix, and W is an orthogonal matrix. Therefore, according to formula (8),

[0108] F'=W'Y(8)

[0109] F'=(f1,f2,...,f q)' is q principal component risk factors. It includes the first n principal component risk factors and qn risk factors, and the first n principal component factors are selected as common factors. Therefore, the W matrix can be expressed as:

[0110] W=(w1,w2,...,w n ,w n+1 ,....,w q )=[W (1) ,W (2) ],

[0111]

[0112] therefore,

[0113] F'=(f1,f2,...,f n ,f n+1 ,...,f q )=[F (1) ,F (2) ],So,

[0114] F (1) =(f1,f2,...,f n ),F (2) =(f n+1 ,f n+2 ,...,f q )

[0115] According to formula (7) and formula (8), we can get:

[0116]

[0117] According to formula (9), assuming δ = W (2) F (2) =(δ1,δ2,...,δ q )',So,

[0118] Y=W (1) F (1) +δ(10), according to the principal component analysis algorithm, the variance of the first n principal component factors is assumed to be:

[0119] λ1,λ2,...,λ n , then assume that:

[0120] According to formula (10), the calculation formula of factor loading matrix B is as follows (11):

[0121] Therefore, the determination of common factors is based on the characteristic root being greater than or equal to 1, and at the same time, the cumulative variance contribution rate of the first n common factors is not less than 85%.

[0122] The original factor loading matrix is ​​B = (b ij ), through orthogonal rotation transformation T, we get the load matrix A = (a ij ).So,

[0123] A=BT, so that the sum of variances of the factor loadings of the common factors reaches the maximum value. The following formula (12) is the sum of variances:

[0124] For the factor loading matrix after orthogonal rotation, The larger it is, the more accurately the information about the top n common risk factors can describe the original information.

[0125] S07: Calculate the inverse matrix of the A matrix, and calculate the correlation coefficient matrix A of the risk factors, such as the inverse matrix of A described in S04.

[0126] S08: Calculate factor scores. The factor score formula is shown in formula (13):

[0127] Where B' is the transpose of the factor loading matrix, A -1 is the inverse matrix of the risk factor correlation coefficient matrix, and Y is defined in step S03. The scores of the common factors are calculated according to formula (13).

[0128] S09: Perform comprehensive evaluation and obtain the comprehensive evaluation value of the common factors by normalizing the weights of the common factors. The calculation is shown in formula (14).

[0129] in, λ i is the variance of the principal component risk factor, as described in step S06.

[0130] like Figure 2 As shown, some risk factors affecting driving behavior in the driving behavior insurance premium rate level determination method of the present invention mainly include three categories: the first category is driver information, the second category is traffic violation information, and the third category is accident data.

[0131] Specifically, the first category of driver information includes static data and dynamic data. The static data includes age, driving experience and gender; the dynamic data includes the number of chronic diseases, the number of times fatigue, panic and palpitations occurred in the past six months, the sleep time and physical condition before driving, participation in safety education and training, etc.

[0132] The second category of traffic violation information includes failure to use turn signals as required, failure to use signal lights, violation of prohibited markings, failure to give way to straight-moving vehicles when turning, pedestrians going first, etc.

[0133] The third category of accident data includes accident vehicle, accident time, accident responsibility, accident cause and number of accidents.

[0134] In some embodiments, the present application also provides a device for determining a driving behavior insurance premium rate level, which can be used to implement the above method, and the device includes:

[0135] Use the vehicle's OBD device to obtain the vehicle's sudden acceleration information, sudden braking information, sharp turn information, and mileage information;

[0136] Use DSM equipment to obtain driver status information, including physical fatigue, number of phone calls, distracted driving, number of smoking, driver abnormal information, and number of times seat belts were not fastened;

[0137] Utilize ADAS equipment to obtain ADAS assisted driving warning information, including the number of times the turn signal is not used in lane change or turning according to regulations, the number of lane deviation warnings, the number of frequent lane change warnings, the number of close vehicle distance and forward collision warnings, and the number of times the traffic light is not used to pass.

[0138] In some embodiments, the present application also provides a system for determining a driving behavior insurance premium rate level, which can be used to implement the above method, and the system includes:

[0139] The data acquisition module uses different acquisition devices to obtain the original data of vehicle driving behavior;

[0140] Data storage module, which stores the collected data on the cloud platform;

[0141] Data processing module, which performs denoising and standardization on raw data;

[0142] Data calculation module, which calculates the standardized data;

[0143] The result display module displays the final evaluation results in different display formats to the corresponding devices through different display programs.

[0144] Example 1

[0145] The data of 2123 cars of a taxi management company for three months were collected for the experiment. According to the above steps S01 and S02, the relevant driving behavior risk factors were calculated, totaling 34 indicators. 34 Represents 34 driving behavior indicators respectively.

[0146] According to the above steps S01 to S09, the following results are obtained:

[0147] (1) The relationship between the common factors F1 and F2 and the 34 indicators is as follows:

[0148] F1=0.806X1+0.742X2+0.403X3+0.511X4+0.102X5+0.302X6+0.293X7+0.820X8

[0149] +0.783X9+0.582X 10 +0.808X 11 +0.912X 12 +0.601X 13 +0.402X 14 +0.809X 15 +0.796X 16

[0150] +0.912X 17 +0.201X 18 +0.300X 19 +0.100X 20 +0.542X 21 +0.801X 22 +0.808X 23 +0.121X 24

[0151] +0.423X 25 +0.652X 26 +0.360X 27 +0.302X 28 +0.206X 29 +0.675X 30 +0.583X 31 +0.992X 32

[0152] +0.653X 33 +0.348X 34

[0153] F2=0.401X1+0.319X2+0.869X3+0.409X4+0.904X5+0.824X6+0.807X7+0.349X8

[0154] +0.529X9+0.648X 10 +0.208X 11 +0.287X 12 +0.547X 13 +0.803X14 +0.342X 15 +0.413X 16

[0155] +0.102X 17 +0.912X 18 +0.846X 19 +0.902X 20 +0.607X 21 +0.267X 22 +0.154X 23 +0.851X 24

[0156] +0.537X 25 +0.304X 26 +0.804X 27 +0.904X 28 +0.645X 29 +0.104X 30 +0.601X 31 +0.007X 32

[0157] +0.485X 33 +0.704X 34

[0158] (2) The first factor has a strong positive correlation with the following indicators, including physiological fatigue (fighting

[0159] yawning, closing eyes), number of phone calls, distracted driving (number of times looking around, number of times looking down),

[0160] Number of times of smoking, abnormal information of the driver (number of times the driver left the driving sight), not wearing a safety belt

[0161] Number of times you drive, driving experience, safety education and training, sleep time before driving, and physical condition

[0162] The correlation coefficients of the nine indicators are 0.806, 0.820, 0.808, 0.912, 0.809,

[0163] 0.912, 0.801, 0.808, 0.992, known as the "static information factor"; the second factor

[0164] The sub-items have a strong positive correlation with the following indicators, including rapid acceleration information, emergency braking

[0165] Vehicle information, sharp turn information, mileage information, lane change, turning without following the regulations

[0166] Number of lights, lane departure warnings, frequent lane change warnings, close distance and forward collision warnings

[0167] Ten indicators, including the number of collision warnings, the number of times traffic lights are not followed, and the number of accidents, are related

[0168] The coefficients are 0.869, 0.904, 0.824, 0.807, 0.912, 0.846, 0.902,

[0169] 0.851, 0.804, and 0.904, which are called “dynamic driving behavior factors”.

[0170] (3) The variance contribution rate and cumulative variance contribution rate are 0.593, 0.438 and 0.985 respectively, indicating that

[0171] Two common factors are sufficient.

[0172] (4) The two common factors can replace 98.5% of the information content of the original data, so the formula is applied

[0173] (15) The rating of vehicle driving behavior insurance premiums can be derived.

[0174] F=(0.593×F1+0.438×F2) / 0.985 (15)

[0175] (5) The rating results of vehicle driving behavior insurance premium rates are shown in Table 1 (the first 10

[0176] Behavior example) as shown:

[0177] Table 1 Results of vehicle insurance grade determination

[0178] Vehicle ID <![CDATA[F1 (First Factor)]]> <![CDATA[F2 (Second Factor)]]> F(comprehensive evaluation) Level determination D202400001 4.1257 2.5864 3.63389 A D202400002 3.0752 2.1045 2.78717 A D202400003 2.9451 1.6925 2.52564 A D202400004 2.5673 1.2347 2.09463 A D202400005 1.5389 1.0459 1.39155 A D202400006 1.0014 1.0001 1.04759 A D202400007 -0.0563 0.6542 0.25701 B D202400008 -0.5794 0.1542 -0.28025 C D202400009 -1.1285 -0.1549 -0.74827 C D202400010 -2.0045 -0.2048 -1.29784 C

[0179] (6) According to the results in Table 1 above, the insurance grades of the ten vehicles are A, B, and C respectively. The vehicles with grade A have a higher risk level of driving behavior, and so on. The lower the grade, the higher the driving safety, and the lowest grade is E. The results in Table 1 above also show that the static information factor of driving behavior and the dynamic driving behavior have a very important influence on the assessment of the insurance grade.

[0180] The present invention greatly expands the existing driving behavior influencing factors through multiple collection methods of driving behavior risk factors, extracts independent common factors using factor models, and extracts common factors with a contribution rate greater than or equal to 85% by calculating the cumulative variance contribution rate of the common factors to describe the driving behavior of the vehicle. Based on the common factors, the vehicle's auto insurance rate is comprehensively evaluated and graded, and verification experiments are carried out using the vehicle's historical insurance policies. The results show that the model accuracy of the present invention is more than 92%, which provides a fair, reasonable and accurate model reference for vehicle auto insurance pricing and promotes the improvement of the overall social welfare.

[0181] It is to be understood that the above embodiments are merely exemplary embodiments used to illustrate the principles of the present invention, but the present invention is not limited thereto. For those of ordinary skill in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. A method for determining the insurance premium rate level of driving behavior, characterized in that: The following steps are involved: S01: Collect driving behavior data; S02: Calculating a driving behavior risk factor value based on the collected driving behavior data; S03: Standardizing the risk factor value; S04: Calculate the correlation coefficient matrix A of the risk factor values; S05: Determine the number of common factors according to the risk factors after the standardization process; S06: Perform orthogonal rotation transformation on the original factor loading matrix; S07: Calculate the inverse matrix of the correlation coefficient matrix A; S08: Calculate factor scores according to the inverse matrix of the correlation coefficient matrix A, the original factor loading matrix and the risk factor matrix; S09: Comprehensively evaluate the common factors according to the factor scores.

2. The method according to claim 1, characterized in that Step S01 includes: Use the vehicle OBD device data to obtain the vehicle's sudden acceleration information, sudden braking information, sharp turn information, and mileage information; Using DSM equipment data, driver status information is obtained, including physical fatigue, number of phone calls, distracted driving, number of smoking, driver abnormal information, and number of times seat belts were not fastened; Utilize ADAS equipment data to obtain ADAS assisted driving warning information, including the number of times the turn signal is not used in lane changes or turns as specified, the number of lane departure warnings, the number of frequent lane change warnings, the number of close vehicle distances and forward collision warnings, and the number of times the vehicle fails to pass according to traffic light instructions.

3. The method according to claim 2, characterized in that Step S02 includes: When calculating the driving behavior risk factor value, the risk factors affecting driving behavior also include driver information, traffic violation data, and accident data, all of which need to be calculated.

4. The method according to claim 3, characterized in that Step S03 includes: The risk factor values ​​are standardized. Assuming that m drivers are evaluated, the driving behavior of each driver includes q behavior indicators, and the constructed matrix Y is: where y i =(y 1i ,y 2i ,...,y mi )', i = 1, 2, ..., q; it needs to be standardized, and the driving behavior risk factor value is standardized using the following formula: in 5. The method according to claim 4, characterized in that Step S04 includes: Calculate the correlation coefficient matrix A of risk factors: Calculate the covariance matrix of q behavioral indicator risk factors, assuming that the correlation coefficient matrix is ​​A = (a ij ) indicates that the calculation is as shown in formula (2): in, Among them, d ij It refers to the standardized transformation of the elements of matrix Y, that is, the standardized processing of the driver's driving behavior indicators; according to the formula, a ij =1, and a ij =a ji .

6. The method according to claim 5, characterized in that Step S05 includes: Determine the number of common factors: Calculate the variance contribution rate and cumulative variance contribution rate of the principal component factor. If the cumulative variance contribution rate reaches 85% or above, then this factor is the principal component factor; Assume that the standardized q-item behavioral indicator risk factor is represented by Y=(y1,y2,…,y q ), independent common factors are represented by F=(F1,F2,…F n ) means, assuming that F1, F2, … F n The mean of is 0 and the variance is 1, so the factor model is expressed by formula (3): Y=B×F(3) in Where B is called the factor loading matrix; For q driving behavior risk factors y i (i=1,2,…q), assuming Represents the sum of the squares of the elements in row i, also the factor y i Since the driving behavior risk factors and common factors have been standardized, and the special factors and common factors are also independent of each other, The larger the value of , the more information contained in the driving behavior risk factor is summarized by n common factors; Assumptions That is, the sum of the squares of the elements in the jth column of the factor loading matrix, expressed as g j represents the sum of the variance contributions of common factors to driving behavior risk factors, g j The larger the value, the more important the common factor j is. Formula (6) represents the variance contribution rate of the common factor: The variance of the risk factor and Where v(y i ) refers to the variance of driving behavior risk factors; The larger the value, the greater the common factor F j The higher the importance.

7. The method according to claim 6, characterized in that Step S06 includes: Original factor loading matrix transformation: In order to obtain more accurate analysis results, it is necessary to transform the original factor loading matrix and calculate the factor loading matrix using the principal component estimation method. The m groups of observations with q behavioral indicators are represented by the matrix Y: Determine n common factors to describe driving behavior; assume that the characteristic matrix W, W satisfies: W'W=WW'=I q (7), Among them I q is the q-order unit matrix, and W is an orthogonal matrix. Therefore, according to formula (8): F'=W'Y(8) F'=(f1,f2,...,f q )' is q principal component risk factors; it includes the first n principal component risk factors and qn risk factors. The first n principal component risk factors are selected as common factors, so the W matrix is ​​expressed as: In=(in1,in2,...,in n ,In n+1 ,....,In q )=[W (1) ,IN (2) ], ， therefore, F'=(f1,f2,...,f n ,f n+1 ,...,f q )=[F (1) ,F (2) ],So, F (1) =(f1,f2,...,f n ),F (2) =(f n+1 ,f n+2 ,...,f q ) According to formula (7) and formula (8), According to formula (9), assuming δ = W (2) F (2) =(δ1,δ2,...,δ q )',So, Y=W (1) F (1) +δ(10), according to the principal component analysis algorithm, the variance of the first n principal component risk factors is assumed to be: λ1,λ2,...,λ n , then assume that: According to formula (10), the calculation formula of factor loading matrix B is as follows (11): Therefore, the determination of common factors is based on the characteristic root being greater than or equal to 1, and at the same time, the cumulative variance contribution rate of the first n common factors is not less than 85%; The original factor loading matrix is ​​B = (b ij ), through orthogonal rotation transformation T, we get the load matrix A = (a ij ),So, A=BT, so that the sum of variances of the factor loadings of the common factors reaches the maximum value. The following formula (12) is the sum of variances: For the factor loading matrix after orthogonal rotation, The larger it is, the more accurately the information about the top n common risk factors can describe the original information.

8. The method according to claim 7, characterized in that Step S08 includes: The factor score formula is shown in formula (13): Among them, B' is the transpose of the factor loading matrix, A -1 is the inverse matrix of the risk factor correlation coefficient matrix, and Y is the definition in step S03; the scores of the common factors are calculated according to formula (13).

9. The method according to claim 8, characterized in that Step S09 includes: The comprehensive evaluation is performed by normalizing the weights of the common factors to obtain the comprehensive evaluation value of the common factors; the calculation is shown in formula (14): in, Among them, λ i is the variance of the principal component risk factors.

10. A device for implementing a method for determining a driving behavior insurance premium rate level according to any one of claims 1 to 9, characterized in that: The device comprises: Use the vehicle's OBD device to obtain the vehicle's sudden acceleration information, sudden braking information, sharp turn information, and mileage information; Use DSM equipment to obtain driver status information, including physical fatigue, number of phone calls, distracted driving, number of smoking, driver abnormal information, and number of times seat belts were not fastened; Utilize ADAS equipment to obtain ADAS assisted driving warning information, including the number of times the turn signal is not used in lane change or turning according to regulations, the number of lane deviation warnings, the number of frequent lane change warnings, the number of close vehicle distance and forward collision warnings, and the number of times the traffic light is not used to pass.

11. A system for implementing a method for determining a driving behavior insurance premium rate level according to any one of claims 1 to 9, characterized in that: The system comprises: The data acquisition module uses different acquisition devices to obtain the original data of vehicle driving behavior; Data storage module, which stores the collected data on the cloud platform; Data processing module, which performs denoising and standardization on raw data; Data calculation module, which calculates the standardized data; The result display module displays the final evaluation results in different display formats to the corresponding devices through different display programs.

Images