Driving behavior evaluation method based on vehicle big data

Abstract

The application discloses a driving behavior evaluation method based on vehicle big data, and the steps include: 1, obtaining vehicle data set from the vehicle big data platform and performing data preprocessing; 2, extracting a driving behavior feature vector by using a principal component analysis method; 3, clustering the driving behavior feature vector by using an improved K-means method, and clustering vehicles with similar driving behavior features into the same cluster; 4, quantitatively evaluating the driving behavior features in three dimensions of fatigue driving, driving behavior safety and driving behavior style, calculating a fatigue driving index and a driving safety index to distinguish the fatigue driving state and the driving behavior safety, and calculating the skewness coefficient and the kurtosis coefficient of acceleration, so that the driving behavior style is classified into a cautious type, a normal type and an aggressive type. The application can classify driving behaviors, quantitatively evaluate the driving behaviors, comprehensively evaluate the safety and risk of the driving behaviors, and improve the driving safety.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of vehicle safety, and in particular to a driving behavior evaluation method based on vehicle big data. BACKGROUND

[0002] With the rapid development of the automotive industry and the progress of intelligentization, more and more vehicles begin to be equipped with various sensors and intelligent devices, which can obtain various data of the vehicle in real time. These data include the speed, acceleration, steering angle, brake force, etc. of the vehicle. At the same time, the popularity of Internet technology also enables vehicles to interact and communicate with other vehicles, traffic facilities and traffic management departments in real time.

[0003] In the prior art, some driving behavior evaluation methods have been proposed and applied. For example, a driving behavior evaluation method based on vehicle sensor data can evaluate the driving behavior safety and risk level of the driver by analyzing parameters such as the acceleration, brake force, and turning speed of the vehicle. In addition, there are some driving behavior evaluation methods based on video images and image processing technology, which can evaluate the attention and reaction ability of the driver by analyzing indicators such as the posture, eye contact and gestures of the driver. However, the existing driving behavior evaluation methods have some problems and shortcomings. First, the traditional evaluation method based on sensor data can only provide some simple driving behavior indicators and cannot comprehensively evaluate the driving behavior of the driver. Second, the evaluation method based on video images requires additional equipment and cost, and there are certain challenges to the privacy protection of the driver. In addition, the existing evaluation methods are subjective and depend on the experience and judgment of experts, lacking objectivity and standardization. SUMMARY

[0004] The present application is to solve the problems existing in the prior art, and proposes a driving behavior evaluation method based on vehicle big data, in order to classify and quantify the driving behavior characteristics, and comprehensively evaluate the safety and risk of driving behavior, so as to improve the driving safety.

[0005] To achieve the above-mentioned application purposes, the present application adopts the following technical solutions:

[0006] The driving behavior evaluation method based on vehicle big data of the present application is characterized by comprising the following steps:

[0007] Step 1, obtaining vehicle data sets of M vehicles from a vehicle big data platform and performing data preprocessing to obtain processed vehicle data sets D={D i |i=1,2,…,M}, wherein D i represents the vehicle data of the i-th vehicle;

[0008] Step 2: Use principal component analysis to extract driving behavior feature vectors F = {f} from the vehicle dataset D. i |i=1,2,…,M}, where f i Let represent the driving behavior feature vector of the i-th vehicle, and Let n represent the nth feature value of the i-th vehicle, where n represents the dimension of the driving behavior feature vector;

[0009] Step 3: Use the improved K-means clustering algorithm to analyze the driving behavior feature vector F = {f} i |i=1,2,…,M} is used to cluster driving behavior, thereby grouping vehicles with similar driving behavior characteristics into the same cluster;

[0010] Step 4: Evaluate the vehicle driving behavior characteristics in different clusters in Step 3.6 using fatigue driving, driving behavior safety, and driving behavior style as three dimensions.

[0011] The indicator for the fatigue driving dimension is the Fatigue Driving Index (FDI), the indicator for the driving behavior safety dimension is the Driving Safety Index (RDSI), and the types of driving behavior style dimensions include: cautious, average, and aggressive.

[0012] Step 4.1: Calculate the Fatigue Driving Index (FDI) of the i-th vehicle. i ;

[0013] when When, it indicates that the driver of the i-th vehicle is in a state of mild fatigue; when When, it indicates that the driver of the i-th vehicle is in a state of moderate fatigue; when When, it indicates that the driver of the i-th vehicle is in a state of severe fatigue; where, These are the lower and upper limits for moderate fatigue, respectively.

[0014] Step 4.2, Calculation of indicators for the driving behavior safety dimension:

[0015] Step 4.2.1: Calculate the driving safety index RDSI of the i-th vehicle using equation (7). i :

[0016]

[0017] In equation (7), ε i，x S represents the weight of the x-th driving behavior indicator for the i-th vehicle. i，x Let x represent the score of the x-th driving behavior indicator for the i-th vehicle, where x = 1, 2, ..., X; and X represent the total number of driving behavior indicators.

[0018] when When, it indicates that the i-th vehicle is in a safe driving state; when , indicates that the i-th vehicle is in a dangerous driving state; when , indicates that the i-th vehicle is in a very dangerous driving state; wherein, are the lower limit value and the upper limit value of the dangerous driving state, respectively;

[0019] Step 4.2.2, calculating the quantitative evaluation value of the safety of the i-th vehicle using formula (10)

[0020]

[0021] In formula (10), r i represents the dangerous state value r i of the i-th vehicle, i R μ represents the total state value;

[0022] , indicates that the driving behavior safety of the i-th vehicle is poor; when , indicates that the driving behavior safety of the i-th vehicle is general; when , indicates that the driving behavior safety of the i-th vehicle is good; wherein, are the lower limit value and the upper limit value of the driving behavior safety, respectively;

[0023] Step 4.3, type determination of the driving behavior style:

[0024] Step 4.3.1, classifying the speed values in the speed data set of the i-th vehicle according to the lower limit threshold value and the upper limit threshold value , thereby dividing into low speed, medium speed, high speed, and recording any one speed as the μ-th speed; μ∈{1, 2, 3} respectively represents low speed, medium speed, and high speed;

[0025] Statistically, the accelerations a μ ={a μι |ι=1,2,...,n μ} under the μ-th speed, wherein a μι represents the i-th acceleration under the μ-th speed, and n μ represents the number of acceleration data sets under the μ-th speed;

[0026] Step 4.3.2, calculating the acceleration skewness coefficients SK={SK μ |μ=1,2,3} and the acceleration kurtosis coefficients KU={KU μ |μ=1,2,3} of the speed data set of the i-th vehicle under the three speeds using formula (11) and formula (12):

[0027]

[0028]

[0029] In formula (11) and formula (12), SK μ represents the acceleration skewness coefficient at the μth speed, KU μ represents the acceleration kurtosis coefficient at the μth speed, represents the average value of the vehicle acceleration at the μth speed, and

[0030] Step 4.3.3, if SK μ ≤α μ , step 4.3.4 is performed; otherwise, step 4.3.5 is performed; wherein α μ represents the set skewness coefficient threshold of the normal driving style at the μth speed;

[0031] Step 4.3.4, if KU μ ≤β μ , the driving style at the μth speed is normal; otherwise, step 4.3.6 is performed; wherein β μ represents the set kurtosis coefficient threshold of the normal driving style at the μth speed;

[0032] Step 4.3.5, if SK μ >α μ , the driving style at the μth speed is aggressive; otherwise, the driving style at the μth speed is cautious;

[0033] Step 4.3.6, if KU μ >β μ , the driving style at the μth speed is aggressive; otherwise, the driving style at the μth speed is cautious.

[0034] The driving behavior evaluation method based on vehicle big data according to the present application is also characterized in that the step 3 comprises the following steps:

[0035] Step 3.1, the optimal clustering number is determined by using the contour coefficient method, and is recorded as K;

[0036] Step 3.2, the driving behavior feature vector f i of the ith vehicle is taken as a to-be-tested sample, and the ith weight factor w i represents the importance of the feature vector f i ;

[0037] The weight factor of the n-dimensional feature value f of the ith vehicle is w , and the weighted vehicle running feature value is recorded as Thus, the new to-be-tested sample of the ith vehicle is obtained and the new driving behavior feature vector F' = {f ′i |i = 1, 2,..., M};

[0038] Step 3.3, clustering F' by using the K-means algorithm to obtain K clusters:

[0039] Step 3.4, obtaining the running feature set of the vehicles in each cluster wherein, denotes the running feature vector of the jth vehicle in the kth cluster, and denotes the n-dimensional feature value of the jth vehicle in the kth cluster, N k denotes the number of vehicles in the kth cluster;

[0040] Step 3.5, calculating the coordinate value of the nth dimension of the kth cluster center by using formula (3) Thus, the updated kth cluster center is obtained and is assigned to

[0041]

[0042] Step 3.6, returning to step b for sequential execution until the cluster center no longer changes, and outputting the final cluster center, so that vehicles with similar driving behavior features are clustered into the same cluster.

[0043] The step 3.3 includes the following steps:

[0044] Step a, randomly initializing the coordinate values of the centers of the K clusters by using formula (1):

[0045]

[0046] In formula (1), denotes the kth cluster center, denotes the coordinate value of the nth dimension of the kth cluster center .

[0047] Step b, calculating the Manhattan distance of the new to-be-tested sample f' of the ith vehicle to the kth cluster center i by using formula (2) Thus, the Manhattan distance of the new to-be-tested sample f' to each cluster center is obtained i

[0048]

[0049] In formula (2), ​​f' represents the new to-be-tested sample of the i-th vehicle i the j-th dimension value of the i-th vehicle, f represents the k-th cluster center the j-th dimension coordinate value of the k-th cluster center;

[0050] Step c, according to the Manhattan distance from the new to-be-tested sample f' to each cluster center, the new to-be-tested sample f' is divided into the nearest cluster; thereby the M vehicle new to-be-tested samples are divided into the nearest cluster, and K clusters are obtained. i i Step c, according to the Manhattan distance from the new to-be-tested sample f' to each cluster center, the new to-be-tested sample f' is divided into the nearest cluster; thereby the M vehicle new to-be-tested samples are divided into the nearest cluster, and K clusters are obtained.

[0051] The fatigue driving index FDI in the step 4.1 i comprises the following steps:

[0052] Step 4.1.1, the driving time T of the i-th vehicle in a day is calculated by using formula (4) i the proportion of the rest time t i η: i，1

[0053]

[0054] Step 4.1.2, the fatigue driving distance is defined as the continuous driving distance exceeding σ kilometers, and the total driving distance D of the i-th vehicle in a day is calculated by using formula (5) i the proportion of the fatigue driving distance d i η: i，2

[0055]

[0056] Step 4.1.3, the fatigue driving index FDI of the i-th vehicle is calculated by using formula (6) i

[0057] FDI i = λ1×η i，1 + λ2×η i，2 (6)

[0058] In formula (6), λ1 and λ2 represent the proportion weight of the fatigue driving time and the proportion weight of the fatigue driving distance respectively.

[0059] In the step 4.2.2, the dangerous state value r i and the total state value R i of the i-th vehicle are calculated by using formula (8) and formula (9) respectively:

[0060]

[0061] ​​​​

[0062] In formula (8), S H represents RDSI i respectively with or Hth intersection point, S H+1 represents RDSI i respectively with or H+1th intersection point, n represents the number of intersection points;

[0063] In formula (9), s0, s end represent the starting and ending mileage of the ith vehicle.

[0064] The electronic device comprises a memory and a processor, and is characterized in that the memory is used for storing a program supporting the processor to execute the driving behavior evaluation method, and the processor is configured to execute the program stored in the memory.

[0065] The computer readable storage medium stores a computer program, and when the computer program is run by a processor, the steps of the driving behavior evaluation method are executed.

[0066] Compared with the prior art, the driving behavior evaluation method has the following beneficial effects:

[0067] 1. The vehicle data analyzed by the driving behavior evaluation method is obtained from a vehicle big data platform, which has a wide coverage range, including various vehicle types and driving scenes, and also collects a large amount of vehicle data. Compared with the traditional data collection method, the vehicle big data platform can obtain more data samples, accurately reflect the general characteristics of driving behavior, and has higher representativeness and objectivity.

[0068] 2. The improved K-means algorithm is used to cluster the driving behavior feature vectors, and a weight factor is introduced for each feature vector to represent its importance and contribution. By introducing the weight factor, the importance of each feature in the driving behavior can be more accurately captured, thereby improving the accuracy and reliability of the clustering results.

[0069] 3. The driving behavior is quantitatively calculated and evaluated from three dimensions of fatigue driving, driving behavior safety, and driving behavior style using vehicle data. Different driving behavior characteristics are represented by objective vehicle data, and the driving style is digitally quantified by calculating the skewness coefficient and kurtosis coefficient of acceleration, which improves the sufficiency and effectiveness of the data, expands the utilization rate of vehicle data, and provides strong data support for drivers, insurance companies, and traffic management departments, thereby promoting traffic safety and improving the comprehensive quality of drivers. BRIEF DESCRIPTION OF DRAWINGS

[0070] Figure 1 An evaluation method flowchart for the present application. DETAILED DESCRIPTION

[0071] In this embodiment, as shown in the figure, a driving behavior evaluation method based on vehicle big data can comprehensively and objectively evaluate driving behavior by using vehicle big data, thereby providing accurate driving behavior information. Specifically, the method comprises the following steps: Figure 1

[0072] Step 1, obtaining vehicle data sets of M vehicles from a vehicle big data platform and performing data preprocessing to obtain processed vehicle data sets D = {D i |i = 1, 2,..., M}, wherein D i represents the vehicle data of the i-th vehicle, and the vehicle data includes vehicle basic information, driving data, location data, driving behavior data, etc. In the preprocessing step, the privacy sensitive data such as vehicle VIN code, location information, license plate number, etc. are deleted, and the redundant, invalid and other segments in the data are processed.

[0073] Step 2, extracting driving behavior feature vectors F = {f i |i = 1, 2,..., M} from the vehicle data set D by using principal component analysis method, wherein f i represents the driving behavior feature vector of the i-th vehicle, and represents the n-dimensional feature value of the i-th vehicle, and n represents the dimension of the driving behavior feature vector.

[0074] Step 3, using an improved K-means clustering algorithm to cluster the driving behavior feature vectors F = {f i |i = 1, 2,..., M};

[0075] Step 3.1, determining the optimal number of clusters by using the silhouette coefficient method, and recording it as K. The clustering result corresponding to the optimal K value has a high silhouette coefficient, indicating that the similarity of the samples is high within the same cluster and low between different clusters.

[0076] Step 3.2, taking the driving behavior feature vector f i of the i-th vehicle as the to-be-tested sample, and letting the i-th weight factor w i represent the importance of the feature vector f i .

[0077] Let the weight factor of the n-dimensional feature value f of the i-th vehicle be w , and the weighted vehicle running feature value is denoted as f . Thus, the new to-be-tested sample of the i-th vehicle is obtained as f And the new driving behavior feature vector F′={f′ i |i = 1, 2, ..., M};

[0078] Step 3.3: Use the K-means algorithm to cluster F′, obtaining K clusters:

[0079] Step a: Randomly initialize the center coordinates of K clusters using equation (1):

[0080]

[0081] In equation (1), Indicates the k-th cluster center. Represents the k-th cluster center The coordinates of the nth dimension;

[0082] Step b: Calculate the new test sample f′ of the i-th vehicle using the Manhattan distance shown in equation (2). i To the k-th cluster center Manhattan distance Thus, the new test sample f′ is obtained. i Manhattan distance to each cluster center:

[0083]

[0084] In equation (2), f′ represents the new test sample for the i-th vehicle. i The j-th eigenvalue in the middle. Represents the k-th cluster center The j-th dimension coordinate value;

[0085] Step c: Based on the new test sample f′ i The distance to Manhattan from each cluster center will affect the new sample f′ to be tested. i The M new test samples are assigned to the nearest cluster, thus creating K clusters.

[0086] Step 3.4: Obtain the set of operational features for vehicles in each cluster. in, Let represent the running feature vector of the j-th vehicle in the k-th cluster, and N represents the nth eigenvalue of the j-th vehicle in the k-th cluster. k This represents the number of vehicles in the k-th cluster;

[0087] Step 3.5: Calculate the coordinates of the nth dimension of the kth cluster center using equation (3). Thus, the updated k-th cluster center is obtained. And assign to

[0088]

[0089] Step 3.6, return to step b sequentially until the cluster center no longer changes, and output the final cluster center, thereby clustering vehicles with similar driving behavior characteristics into the same cluster;

[0090] Step 4, taking fatigue driving, driving behavior safety, and driving behavior style as three dimensions, the driving behavior characteristics of vehicles in different clusters of step 3.6 are evaluated;

[0091] Let the index of the fatigue driving dimension be fatigue driving index FDI, the index of the driving behavior safety dimension be driving safety index RDSI, and the types of the driving behavior style dimension include cautious type, ordinary type, and aggressive type;

[0092] Step 4.1, index calculation of the fatigue driving dimension:

[0093] Step 4.1.1, calculate the driving time T of the ith vehicle on a certain day using formula (4) i The proportion of rest time t i η i，1 :

[0094]

[0095] Step 4.1.2, define the fatigue driving distance as the continuous driving distance exceeding σ kilometers, and calculate the total driving distance D of the ith vehicle on a certain day using formula (5) i The proportion of fatigue driving distance d i η i，2 :

[0096]

[0097] Step 4.1.3, calculate the fatigue driving index FDI of the ith vehicle using formula (6) i :

[0098] FDI i = λ1 × η i，1 + λ2 × η i，2 (6)

[0099] In formula (6), λ1 and λ2 represent the proportion weight of fatigue driving time and the proportion weight of fatigue driving distance, respectively;

[0100] Step 4.1.4, when , it indicates that the driver of the ith vehicle is in a mild fatigue state; when , it indicates that the driver of the ith vehicle is in a moderate fatigue state; when , indicates that the driver of the ith vehicle is in a serious fatigue state; wherein, respectively are the lower limit value and the upper limit value of the moderate fatigue state;

[0101] Step 4.2, index calculation of the driving behavior safety dimension:

[0102] Step 4.2.1, calculate the driving safety index RDSI of the ith vehicle using formula (7) i :

[0103]

[0104] In formula (7), ε i，x represents the weight of the xth driving behavior index of the ith vehicle, S i，x represents the score of the xth driving behavior index of the ith vehicle, x = 1, 2, …, X; X represents the total number of driving behavior indexes;

[0105] The driving behavior index can be derived from the longitudinal motion, the lateral motion, the driving smoothness, and the fatigue driving. The parameters of each driving behavior index are shown in Table 1:

[0106] Table 1 Driving behavior index table

[0107]

[0108] The CRITIC method is used to determine the index weight. The weight of the xth driving behavior index of the ith vehicle is calculated using formula (8) i，x :

[0109]

[0110] In formula (8), δ i，x represents the standard deviation of the xth driving behavior index of the ith vehicle, represents the correlation coefficient of the xth driving behavior index and the yth driving behavior index of the ith vehicle;

[0111] The driving behavior index score function is designed. The index score rule is formulated, the 0 score threshold is set, and the index score function is converted into a piecewise function of 0 and a linear function:

[0112]

[0113] In formula (9), c i，x represents the parameter value of the xth driving behavior index of the ith vehicle, τ i，x represents the parameter threshold of the xth driving behavior index of the ith vehicle;

[0114] When When, it indicates that the i-th vehicle is in a safe driving state; when When, it indicates that the i-th vehicle is in a dangerous driving state; when When, it indicates that the i-th vehicle is in an extremely dangerous driving state; where, These are the lower and upper limits for dangerous driving conditions, respectively.

[0115] Step 4.2.2: Based on the mileage s of the i-th vehicle... i and Driving Safety Index (RDSI) i The danger state value r of the i-th vehicle is calculated using equations (8) and (9) respectively. i and total state value R i :

[0116]

[0117]

[0118] Equation (10), S H Indicates RDSI i respectively with or The Hth intersection point, S H+1 Indicates RDSI i respectively with or The (H+1)th intersection point, where n represents the number of intersection points;

[0119] Equation (11), s0, s end Indicates the starting and ending mileage of the i-th vehicle;

[0120] Step 4.2.3: Calculate the quantitative evaluation value of the safety of the i-th vehicle using equation (12).

[0121]

[0122] when When, it indicates that the driving behavior of the i-th vehicle is less safe; when When, it indicates that the driving safety of the i-th vehicle is generally acceptable; when When, it indicates that the driving behavior of the i-th vehicle is relatively safe; among which, These are the lower and upper limits for driving behavior safety, respectively;

[0123] Step 4.3, Determining the Type of Driving Behavior Style:

[0124] Step 4.3.1: Based on the lower limit threshold of medium speed and upper limit threshold Classify the speed values in the speed data set of the ith vehicle, so as to be divided into low speed, medium speed, high speed, and any one speed is recorded as the μth speed; μ∈{1, 2, 3} respectively represent low speed, medium speed, high speed;

[0125] Statistics of the acceleration a μ at the μth speed μι |ι=1,2,...,n μ}, wherein a μι represents the ith acceleration at the μth speed, n μ represents the number of acceleration data sets at the μth speed;

[0126] Step 4.3.2, calculate the acceleration skewness coefficient SK = {SK μ |μ=1,2,3} and the acceleration kurtosis coefficient KU = {KU μ |μ=1,2,3} of the speed data set of the ith vehicle at the three speeds by using formula (13) and formula (14):

[0127]

[0128]

[0129] In formula (13) and formula (14), SK μ represents the acceleration skewness coefficient at the μth speed, KU μ represents the acceleration kurtosis coefficient at the μth speed, represents the average value of the vehicle acceleration at the μth speed, and

[0130] Skewness is a measure of the direction and degree of skewness of statistical data distribution, and is a numerical feature of the degree of asymmetry of statistical data distribution. Skewness of 0 is normal distribution, skewness greater than 0 is positive skewness, and skewness less than 0 is negative skewness; kurtosis is a characteristic number of the peak value of the probability density distribution curve at the average value, the greater the kurtosis coefficient, the more extreme values the distribution has, and the smaller the kurtosis coefficient, the more the distribution is near the mean value;

[0131] Step 4.3.3, if SK μ ≤α μ , then step 4.3.4 is executed; otherwise, step 4.3.5 is executed; wherein α μ represents the skewness coefficient threshold of the ordinary driving style at the μth speed set;

[0132] Step 4.3.4, if KU μ ≤β μ , then the driving style at the μth speed is ordinary; otherwise, step 4.3.6 is executed; wherein βμ a kurtosis coefficient threshold value of the normal driving style at the μth speed set;

[0133] Step 4.3.5, if SK μ > a μ , it indicates that the driving style at the μth speed is aggressive; otherwise, it indicates that the driving style at the μth speed is cautious;

[0134] Step 4.3.6, if KU μ > b μ , it indicates that the driving style at the μth speed is aggressive; otherwise, it indicates that the driving style at the μth speed is cautious.

[0135] In this embodiment, an electronic device includes a memory for storing a program supporting a processor to execute the above method, and the processor is configured to execute the program stored in the memory.

[0136] In this embodiment, a computer readable storage medium has a computer program stored thereon, and the computer program is executed by a processor to perform the steps of the above method.

Claims

1. A driving behavior evaluation method based on vehicle big data, characterized in that, The method comprises the following steps: Step 1, obtaining vehicle data set of the vehicle from the vehicle big data platform and performing data preprocessing to obtain processed vehicle data set , wherein, represents the vehicle data of the i-th vehicle ​​ Step 2: Use principal component analysis to analyze the vehicle dataset. Extract driving behavior feature vectors ,in, Indicates the first The driving behavior feature vector of the vehicle, and , Indicates the first The first car 3D eigenvalues The dimension representing the feature vector of driving behavior; Step 3, using improved K-means clustering algorithm to the driving behavior feature vector driving behavior clustering, so that the driving behavior similar to the vehicle as a cluster of the same cluster; Step 4, taking fatigue driving, driving behavior safety and driving behavior style as three dimensions, evaluating the driving behavior characteristics of vehicles in different clusters; Let the index of fatigue driving dimension be fatigue driving index , the index of driving behavior safety dimension be driving safety index , the type of driving behavior style dimension include: cautious type, ordinary type, aggressive type; Step 4.1, calculating a fatigue driving index for a vehicle ;​ when When, it indicates the first The driver of the vehicle was in a state of mild fatigue; when When, it indicates the first The driver of the vehicle was in a state of moderate fatigue; when When, it indicates the first The driver of the vehicle was severely fatigued; among them, , These are the lower and upper limits for moderate fatigue, respectively. Step 4.2, index calculation of the driving behavior safety dimension: Step 4.2.1, calculating the driving safety index for a vehicle using formula (7) vehicle's driving safety index : (7) In formula (7), represents the weight of the first driving behavior index of the vehicle, represents the score of the first driving behavior index of the vehicle, ; X represents the total number of driving behavior indexes.​​ when When, it indicates the first The vehicle is in a safe driving condition; when When, it indicates the first The vehicle is in a dangerous driving state; when When, it indicates the first The vehicle was in an extremely dangerous driving condition; among them... , These are the lower and upper limits for dangerous driving conditions, respectively. Step 4.2.

2. Calculating the quantified evaluation value of the safety of the vehicle using Equation (10) :​ (10) In formula (10), represents the total state value; and represents the total state value; and , represents the total state value; and when When, it indicates the first The driving behavior of the vehicle is less safe; when When, it indicates the first The vehicle's driving safety is generally low; when When, it indicates the first The driving behavior of the vehicle is relatively safe; among them, , These are the lower and upper limits for driving behavior safety, respectively; Step 4.3, type determination of the driving behavior style: Step 4.3.1: Based on the lower limit threshold of medium speed and upper limit threshold , for the The speed values ​​in the vehicle speed dataset are classified into low speed, medium speed, and high speed, and any speed is denoted as the [number]. This kind of speed; These represent low speed, medium speed, and high speed, respectively. Statistics in the first acceleration at a certain velocity ,in, Indicates the first The first speed An acceleration, Indicates the first The number of acceleration datasets at various velocities; Step 4.3.2, calculating the skewness and kurtosis of the acceleration data set of the vehicle at the three speeds using equations (11) and (12) :​​ (11) (12) In formula (11) and formula (12), indicates the acceleration skewness coefficient at the indicates the acceleration kurtosis coefficient at the indicates the average value of the vehicle acceleration at the ;​​​ Step 4.3.3, if Step 4.3.4 is performed; otherwise, Step 4.3.5 is performed; wherein, denotes a set of speed-dependent skewness coefficient thresholds for the normal driving style; and denotes a set of speed-dependent skewness coefficient thresholds for the normal driving style. Step 4.3.4, if , indicates that the driving style at the th speed is of the ordinary type; otherwise, Step 4.3.6 is performed; wherein, indicates the kurtosis coefficient threshold value of the ordinary type driving style at the th speed set. Step 4.3.5, if indicates that the driving style at the th speed is aggressive; otherwise, it indicates that the driving style at the th speed is cautious. Step 4.3.6, if indicates that the driving style at the th speed is aggressive; otherwise, it indicates that the driving style at the th speed is cautious.

2. The driving behavior evaluation method based on vehicle big data according to claim 1, characterized in that, The step 3 comprises the following steps: Step 3.1, determine the optimal number of clusters using silhouette coefficient method and denote it as ; Step 3.2, place the first Vehicle driving behavior feature vector As the sample to be tested, let the first... Weighting factors Representing the eigenvector The importance of; Order No. The first car 3D eigenvalues The weighting factor is The weighted vehicle operating characteristic value is denoted as Thus, the first New test samples of vehicles And new driving behavior feature vectors ; Step 3.

3. Clustering the data using K-means algorithm to get clusters: clusters: Step 3.4: Obtain the set of operational features for vehicles in each cluster. ,in, Indicates the first The th cluster The vehicle's operational feature vector, and , Indicates the first The th cluster The first car 3D eigenvalues Indicates the first The number of vehicles in each cluster; Step 3.5, calculate the coordinate value of the first dimension of the first cluster center using formula (3) , , , , , , ; (3) Step 3.6, sequentially performing step 3.3 until the cluster center no longer changes, and outputting the final cluster center, so that vehicles with similar driving behavior characteristics are clustered into the same cluster.

3. The driving behavior evaluation method based on vehicle big data according to claim 2, characterized in that, The step 3.3 comprises the following steps: Step a, random initialization with formula (1) The center coordinates of the cluster: (1) In formula (1), denotes the cluster center, denotes the cluster center a coordinate value of the dimension of the Step b, calculating Manhattan distances of the new to-be-tested sample of the vehicle from the cluster centers of the first to the th cluster center using the Manhattan distance shown in formula (2) , so as to obtain the Manhattan distances of the new to-be-tested sample of the vehicle from the cluster centers of the first to the th cluster center . ​ (2) In formula (2), represent the new to-be-tested sample of the vehicle represent the new to-be-tested sample of the vehicle represent the new to-be-tested sample of the vehicle represent the new to-be-tested sample of the vehicle represent the new to-be-tested sample of the vehicle represent the new to-be-tested sample of the vehicle represent the new to-be-tested sample of the vehicle​ Step c, the new sample to be tested is divided into the cluster closest to it according to the Manhattan distance to the cluster center Step c, the new sample to be tested is divided into the cluster closest to it according to the Manhattan distance to the cluster center Step c, the new sample to be tested is divided into the cluster closest to it according to the Manhattan distance to the cluster center Step c, the new sample to be tested is divided into the cluster closest to it according to the Manhattan distance to the cluster center Step c, the new sample to be tested is divided into the cluster closest to it according to the Manhattan distance to the cluster center 4. The driving behavior evaluation method based on vehicle big data according to claim 3, characterized in that, the fatigue driving index in step 4.1 the calculation comprises the steps of: Step 4.1.

1. Calculate the percentage of the day that the vehicle is driven :​​​ (4) Step 4.1.2, defining the continuous driving over kilometers as the fatigue driving distance, the ratio of the fatigue driving distance to the total driving distance of the vehicle on a certain day is calculated using equation (5) :​​​ (5) Step 4.1.3, calculating the fatigue driving index for a vehicle using formula (6) vehicle's fatigue driving index : (6) In formula (6), respectively represent the proportional weight of the fatigue driving time, the proportional weight of the fatigue driving distance.

5. The driving behavior evaluation method based on vehicle big data according to claim 2, characterized in that, The steps 4.2.2 are calculated in the formula (8) and formula (9) respectively dangerous state value of the vehicle and total state value : (8) (9) In formula (8), denotes respectively with or the first intersection, denotes respectively with or the first intersection, n denotes the number of intersections; In equation (9), , Indicates the first The starting and ending mileage of the vehicle.

6. An electronic device comprising a memory and a processor, characterized in that The memory is configured to store a program supporting the processor to execute the driving behavior evaluation method in any one of claims 1-5, and the processor is configured to execute the program stored in the memory.

7. A computer-readable storage medium having stored thereon a computer program, characterized in that The computer program is executed by the processor to perform the steps of the driving behavior evaluation method in any one of claims 1-5.

Images