CN106022926A - Premium deduction method based on mileage and driving behavior safety - Google Patents

Abstract

The invention belongs to the technical field of vehicle insurance, and specifically relates to a premium deduction method based on mileage and driving behavior safety. The method comprises the following steps of: (1) collecting data sources; (2) initially setting variables expressing driving behavior risk factors according to the data sources, using a Logistic regression model to modeling the variables, and calculating a risk probability for evaluating the driving behavior safety according to a probability solving method of the Logistic regression model; (3) calculating a safety score and a safe driving coefficient per kilometer; and (4) constructing a premium deduction model. According to the invention, the usage behavior is recorded, modeling analysis is carried out and the driving behavior is quantified, a safety scoring system is constructed based on the behavior science, and the fairness and scientificity of premium pricing and deduction are substantially improved.

Description

A kind of premium based on mileage and driving behavior safety reduces method

Technical field

The invention belongs to car insurance technical field, be specifically related to a kind of based on mileage with the premium of driving behavior safety Reduce method.

Background technology

In the evolution of car insurance, pricing model always promotes the key factor of car insurance progress.China The most substantially fix a price the stage still in protection amount, and to half insured amount, half vehicle price transition.Either insured amount price or car Type fix a price, in the pricing model that these are traditional, Pricing Factor include from car, from ground, from with from factors such as people.Wherein, " from With " refer to use classes, it is i.e. business car or homebrew car；" from people " is the factor static concept, such as age, property Not, marriage and occupation etc..These Pricing Factors reflect risk difference to a certain extent, but still can not objectively reflect wind Danger situation, because " risk based on using " is the most important risk factor of car insurance, as driving behavior is accustomed to；Even The people that age, sex, marriage are the most identical with occupation, the driving behavior custom between them there may be a world of difference, rate of being in danger May differ widely.Therefore, traditional insured amount price and vehicle price this " static schema " all face science and fairness Challenge.There is the defect that accuracy science low, inadequate is fair in above-mentioned prior art.

Car networking and the appearance of big data technique, bring possibility to cracking tradition pricing model limitation.Propose in the present invention " premium based on mileage and driving behavior safety reduces method " pricing model in, " with " and " people " more refer to dynamically Concept, based on individual actually used situation, including distance travelled, time, region and driving behavior custom etc., wherein drives row For being the emphasis considered.

Summary of the invention

It is an object of the invention to for the deficiencies in the prior art, it is provided that a kind of accuracy height, more science and the base of justice Premium in mileage and driving behavior safety reduces method.

The present invention solves the technical scheme of problem: a kind of premium side of reducing based on mileage and driving behavior safety Method, comprises the steps:

(1) gather data source, described data source include the information of driver, information of vehicles, Weather information, road information, Traffic information, vehicle sensor data；

(2) variable of driving behavior risks and assumptions is represented according to data source initial setting, with Logistic regression model pair Variable is modeled, and asks the method for probability to obtain the risk for evaluating driving behavior safety according to Logistic regression model Probability；

(3) safety and the safe driving coefficient of every kilometer are obtained:

The computing formula of safety is:

S_i=100-(p_i-0.5) × 20,

Wherein, S_iRepresent the safety of i-th kilometer, p_iRepresent the risk probability of i-th kilometer, i=1,2 ... n, n represent The milimeter number travelled；

According to the safety of every kilometer, show that every kilometer of corresponding safe driving coefficient, safety and safety are driven The corresponding relation sailing coefficient is as follows:

Safety is [96～100], [90～95], [80～89], [60～79], [40～59], [20～39], [0～ 19] corresponding safe driving coefficient is respectively 0.5,0.75,1,1.5,2,3,5；

(4) build premium and reduce model: described premium is reduced the formula of model and is:

$B=IP\×60\%-{\mathrm{\Σ}}_{i=0}^{n}U\×C_i,$

Wherein, B represents premium remaining sum, and IP represents premium total amount, C_iRepresenting the safe driving coefficient of i-th kilometer, U represents Every kilometer of foundation unit price, the computing formula of U is:

$U=\frac{IP\×60\%}{D},$

Wherein, D represents the year distance travelled number of the vehicle that vehicle is corresponding, and described D is by obtaining the vehicle product of described vehicle Board, the year driving mileage data storehouse of enquiring vehicle brand obtains.

Further, described step (2) comprises the steps:

(2.1) variable of driving behavior risks and assumptions is represented according to data source initial setting；

(2.2) with Logistic regression model, variable data is modeled；

(2.3) use AIC information criterion (Akaike Information Criterion, akaike information criterion) right The regression result of Logistic regression model carries out Variable Selection, and the number of the variable filtered out is k；

(2.4) based on k the variable filtered out, calculating the risk probability of every kilometer, computing formula is:

$p_i=\frac{e^{{\mathrm{\β}}_0+{\mathrm{\β}}_1{x}_1+...+{\mathrm{\β}}_k{x}_k}}{1+e^{{\mathrm{\β}}_0+{\mathrm{\β}}_1{x}_1+...+{\mathrm{\β}}_k{x}_k}},$

Wherein, p_iRepresent the risk probability of i-th kilometer, x_kRepresent kth variable, β_kRepresent that kth variable is at i-th kilometer Regression coefficient.

Further, in described step (2.1), the number of the variable of initial setting is 96, including 1 dependent variable and 95 Individual independent variable.

Further, in described step (2.3), described in the number k=13 of variable that filters out.

Further, the variable filtered out described in 13 be year accumulative mileage, accumulation trip natural law, daily go out line number, Average instantaneous oil consumption, average direction dish corner, the absolute value of average negative sense steering wheel angle, speed per hour standard deviation, steering wheel Corner standard deviation, acceleration standard deviation, the absolute value of minimum acceleration, hundred kilometers of anxious deceleration number of times, night trip durations account for always Travel time ratio, more than 90km/h speed travel time accounting.

Further, in described step (1), described vehicle sensor data is by installing vehicle-mounted T-BOX before vehicle (Telematics BOX, remote information processor) equipment or after vehicle install OBD (On-Board Diagnostic, vehicle-mounted Diagnostic system) equipment acquisition.

Further, in described vehicle sensor data includes Vehicle Identify Number, data receipt time, longitude, latitude, kilometer Journey, the tire pressure of front revolver, the tire pressure of front right wheel, the tire pressure of rear revolver, the tire pressure of rear right wheel, instantaneous oil consumption, longitudinal acceleration, with Corner metering the position of steering wheel, the rotating speed of steering wheel, speed, engine speed, and front left-hand door, front right door, rear left-hand door, Rear right door, boot, the state of steering wheel sensor, described state is for opening or closing.

Preferably, in described step (1), the information of described driver includes: sex, age, education degree, wedding are no, receive Enter, occupation, the driving age, place of working, residence；

Described information of vehicles includes: vehicle, car age, hands automatically, fuel type, car fare, car color；

Described Weather information includes: normal, light rain, moderate rain, heavy rain, slight snow, moderate snow, heavy snow, mild wind, strong wind, hurricane, Mist, haze；

Described road information includes: at a high speed, national highway, provincial highway, county road, township road, village of counties and townships internal road, city free way, master Want road, secondary road, ordinary road, path；

Described traffic information includes: unimpeded, walk or drive slowly, block up, wagon flow average speed.

Preferably, in described step (2.1), the variable of 96 described initial settings is as shown in the table:

The invention have the benefit that the method for the invention passes through induction system, record usage behavior data, modeling point Analyse and quantify driving behavior, build the safe score-system of Behavior-based control science, significantly improve the price of premium and reduce Fairness, science.Meanwhile, by risks and assumptions is selected, specify that the risks and assumptions with risk height correlation and each From weight, the risk probability for each section of driving performance of scientific evaluation provides foundation.

Accompanying drawing explanation

Fig. 1 is the flow chart that premium based on mileage and driving behavior safety of the present invention reduces method.

Detailed description of the invention

Below in conjunction with the accompanying drawings and detailed description of the invention, the present invention is further illustrated.

As it is shown in figure 1, a kind of premium based on mileage and driving behavior safety reduces method, comprise the steps:

(1) gather data source, described data source include the information of driver, information of vehicles, Weather information, road information, Traffic information, vehicle sensor data；

(2) variable of driving behavior risks and assumptions is represented according to data source initial setting, with Logistic regression model pair Variable is modeled, and asks the method for probability to obtain the risk for evaluating driving behavior safety according to Logistic regression model Probability；

(3) safety and the safe driving coefficient of every kilometer are obtained:

The computing formula of safety is:

S_i=100-(p_i-0.5) × 20,

Wherein, S_iRepresent the safety of i-th kilometer, p_iRepresent the risk probability of i-th kilometer, i=1,2 ... n, n represent The milimeter number travelled；

According to the safety of every kilometer, show that every kilometer of corresponding safe driving coefficient, safety and safety are driven The corresponding relation sailing coefficient is as follows:

Safety is [96～100], [90～95], [80～89], [60～79], [40～59], [20～39], [0～ 19] corresponding safe driving coefficient is respectively 0.5,0.75,1,1.5,2,3,5；

(4) build premium and reduce model: described premium is reduced the formula of model and is:

$B=IP\×60\%-{\mathrm{\Σ}}_{i=0}^{n}U\×C_i,$

Wherein, B represents premium remaining sum, and IP represents premium total amount, C_iRepresenting the safe driving coefficient of i-th kilometer, U represents Every kilometer of foundation unit price, the computing formula of U is:

$U=\frac{IP\×60\%}{D},$

Wherein, D represents the year distance travelled number of the vehicle that vehicle is corresponding, and described D is by obtaining the vehicle product of described vehicle Board, the year driving mileage data storehouse of enquiring vehicle brand obtains.

Described step (2) comprises the steps:

(2.1) variable of driving behavior risks and assumptions is represented according to data source initial setting；

(2.2) with Logistic regression model, variable data is modeled；

(2.3) use AIC information criterion that the regression result of Logistic regression model is carried out Variable Selection, filter out The number of variable is k；

(2.4) based on k the variable filtered out, calculating the risk probability of every kilometer, computing formula is:

$p_i=\frac{e^{{\mathrm{\β}}_0+{\mathrm{\β}}_1{x}_1+...+{\mathrm{\β}}_k{x}_k}}{1+e^{{\mathrm{\β}}_0+{\mathrm{\β}}_1{x}_1+...+{\mathrm{\β}}_k{x}_k}},$

Wherein, p_iRepresent the risk probability of i-th kilometer, x_kRepresent kth variable, β_kRepresent that kth variable is at i-th kilometer Regression coefficient.

In described step (2.1), the number of the variable of initial setting is 96, including 1 dependent variable and 95 independent variables.

In described step (2.3), described in the number k=13 of variable that filters out.

The variable filtered out described in 13 be year accumulative mileage, accumulation trip natural law, daily go out line number, average instantaneous oil Consumption, average direction dish corner, the absolute value of average negative sense steering wheel angle, speed per hour standard deviation, steering wheel angle standard deviation, Acceleration standard deviation, the absolute value of minimum acceleration, hundred kilometers of anxious deceleration number of times, night trip durations account for ratio of total travel time Example, more than 90km/h speed travel time accounting.

In described step (1), described vehicle sensor data is by installing T-box equipment or pacifying after vehicle before vehicle Dress OBD equipment obtains.

Described vehicle sensor data includes Vehicle Identify Number, data receipt time, longitude, latitude, kilometrage, front revolver Tire pressure, the tire pressure of front right wheel, the tire pressure of rear revolver, the tire pressure of rear right wheel, instantaneous oil consumption, longitudinal acceleration, with corner metering The position of steering wheel, the rotating speed of steering wheel, speed, engine speed, and front left-hand door, front right door, rear left-hand door, rear right door, after Standby case, the state of steering wheel sensor, described state is for opening or closing.

Embodiment one

A kind of premium based on mileage and driving behavior safety reduces method, comprises the steps:

(1) gather data source, described data source include the information of driver, information of vehicles, Weather information, road information, Traffic information, vehicle sensor data；

The information of described driver includes: sex, age, education degree, wedding are no, income, occupation, the driving age, place of working, residence Residence；

Described information of vehicles includes: vehicle, car age, hands automatically, fuel type, car fare, car color；

Described Weather information includes: normal, light rain, moderate rain, heavy rain, slight snow, moderate snow, heavy snow, mild wind, strong wind, hurricane, Mist, haze；

Described road information includes: at a high speed, national highway, provincial highway, county road, township road, village of counties and townships internal road, city free way, master Want road, secondary road, ordinary road, path；

Described traffic information includes: unimpeded, walk or drive slowly, block up, wagon flow average speed.

Described vehicle sensor data includes Vehicle Identify Number, data receipt time, longitude, latitude, kilometrage, front revolver Tire pressure, the tire pressure of front right wheel, the tire pressure of rear revolver, the tire pressure of rear right wheel, instantaneous oil consumption, longitudinal acceleration, with corner metering The position of steering wheel, the rotating speed of steering wheel, speed, engine speed, and front left-hand door, front right door, rear left-hand door, rear right door, after Standby case, the respective state of steering wheel sensor, described state is for opening or closing.

(Telematics BOX, at remote information by installing vehicle-mounted T-BOX before vehicle for described vehicle sensor data Reason device) equipment or after vehicle install OBD (On-Board Diagnostic, onboard diagnostic system) equipment obtain；

(2) ask for risk probability, specifically include following steps:

(2.1) variable of driving behavior risks and assumptions is represented according to data source initial setting 96；96 described tentatively sets Fixed variable is as shown in the table:

(2.2) with Logistic regression model, variable data is modeled；

(2.3) use AIC information criterion that the regression result of Logistic regression model is carried out Variable Selection, filter out The number of variable is 13, specific as follows shown:

(2.4) based on 13 variablees filtered out, calculating the risk probability of every kilometer, computing formula is:

$p_i=\frac{e^{{\mathrm{\β}}_0+{\mathrm{\β}}_1{x}_1+...+{\mathrm{\β}}_{k13}{x}_{13}}}{1+e^{{\mathrm{\β}}_0+{\mathrm{\β}}_1{x}_1+...+{\mathrm{\β}}_{13}{x}_{k13}}},$

Wherein, p_iRepresent the risk probability of i-th kilometer, x₁₃Represent the 13rd variable, β₁₃Represent that the 13rd variable is public i-th In regression coefficient.Wherein, in above formula:

β₀+β₁x₁+…+β₁₃x₁₃=0+cumdays_year*0.427+cumdays_year*0.254+tripday*- 0.282+mean_consum*-0.154+mean_psteer*-0.287+mean_nsteer*-0.296+sd_speed*- 0.142+sd_steer*0.421+sd_acce*-0.234+accemin*0.135+accen3_per100*0.386+night_ ratio*0.082+speed90_ratio*-0.182；

In the risk probability calculated as above table shown in p value string；

(3) safety and the safe driving coefficient of every kilometer are obtained:

The computing formula of safety is:

S_i=100-(p_i-0.5) × 20,

Wherein, S_iRepresent the safety of i-th kilometer, p_iRepresent the risk probability of i-th kilometer, i=1,2 ... n, n represent The milimeter number travelled；

According to the safety of every kilometer, show that every kilometer of corresponding safe driving coefficient, safety and safety are driven The corresponding relation sailing coefficient is as follows:

Safe driving coefficient	Safety
		0.5	[96-100]
0.75	[90-95]
		1	[80-89]
1.5	[60-79]
		2	[40-59]
3	[20-39]
		5	[0-19]

(4) build premium and reduce model: described premium is reduced the formula of model and is:

$B=IP\×60\%-{\mathrm{\Σ}}_{i=0}^{n}U\×C_i,$

Wherein, B represents premium remaining sum, and IP represents premium total amount, C_iRepresenting the safe driving coefficient of i-th kilometer, U represents Every kilometer of foundation unit price, the computing formula of U is:

$U=\frac{IP\×60\%}{D},$

Wherein, D represents the year distance travelled number of the vehicle that vehicle is corresponding, and described D is by obtaining the vehicle product of described vehicle Board, the year driving mileage data storehouse of enquiring vehicle brand obtains.

The present invention is not limited to above-mentioned embodiment, in the case of without departing substantially from flesh and blood of the present invention, and art technology Personnel it is contemplated that any deformation, improve, replace and each fall within protection scope of the present invention.

Claims (7)

1. a premium based on mileage and driving behavior safety reduces method, it is characterised in that comprise the steps:

(1) gathering data source, described data source includes the information of driver, information of vehicles, Weather information, road information, road conditions Information, vehicle sensor data；

(2) variable of driving behavior risks and assumptions is represented according to data source initial setting, with Logistic regression model to variable It is modeled, general according to the risk that Logistic regression model asks the method for probability to obtain for evaluating driving behavior safety Rate；

(3) safety and the safe driving coefficient of every kilometer are obtained:

The computing formula of safety is:

S_i=100-(p_i-0.5) × 20,

Wherein, S_iRepresent the safety of i-th kilometer, p_iRepresent the risk probability of i-th kilometer, i=1,2 ... n, n represent traveling Milimeter number；

According to the safety of every kilometer, draw every kilometer of corresponding safe driving coefficient, safety and safe driving system The corresponding relation of number is as follows:

Safety is [96～100], [90～95], [80～89], [60～79], [40～59], [20～39], [0～19] Corresponding safe driving coefficient is respectively 0.5,0.75,1,1.5,2,3,5；

(4) build premium and reduce model: described premium is reduced the formula of model and is:

$B=IP\×60\%-{\mathrm{\Σ}}_{i=0}^{n}U\×C_i,$

Wherein, B represents premium remaining sum, and IP represents premium total amount, C_iRepresenting the safe driving coefficient of i-th kilometer, U represents every public affairs In foundation unit price, the computing formula of U is:

$U=\frac{IP\×60\%}{D},$

Wherein, D represents the year distance travelled number of the vehicle that vehicle is corresponding, and described D, by obtaining the vehicle brand of described vehicle, looks into The year driving mileage data storehouse asking vehicle brand obtains.

Premium based on mileage and driving behavior safety the most according to claim 1 reduces method, it is characterised in that institute State step (2) to comprise the steps:

(2.1) variable of driving behavior risks and assumptions is represented according to data source initial setting；

(2.2) with Logistic regression model, variable data is modeled；

(2.3) use AIC information criterion that the regression result of Logistic regression model is carried out Variable Selection, the variable filtered out Number be k；

(2.4) based on k the variable filtered out, calculating the risk probability of every kilometer, computing formula is:

$p_i=\frac{e^{{\mathrm{\β}}_0+{\mathrm{\β}}_1{x}_1+...+{\mathrm{\β}}_k{x}_k}}{1+e^{{\mathrm{\β}}_0+{\mathrm{\β}}_1{x}_1+...+{\mathrm{\β}}_k{x}_k}},$

Wherein, p_iRepresent the risk probability of i-th kilometer, x_kRepresent kth variable, β_kRepresent kth variable i-th kilometer return Return coefficient.

Premium based on mileage and driving behavior safety the most according to claim 2 reduces method, it is characterised in that institute Stating in step (2.1), the number of the variable of initial setting is 96, including 1 dependent variable and 95 independent variables.

Premium based on mileage and driving behavior safety the most according to claim 2 reduces method, it is characterised in that institute State in step (2.3), described in the number k=13 of variable that filters out.

Premium based on mileage and driving behavior safety the most according to claim 4 reduces method, it is characterised in that 13 The variable filtered out described in individual be year accumulative mileage, accumulation trip natural law, daily go out line number, average instantaneous oil consumption, average the most just To steering wheel angle, the absolute value of average negative sense steering wheel angle, speed per hour standard deviation, steering wheel angle standard deviation, accelerate scale Accurate poor, the absolute value of minimum acceleration, hundred kilometers of anxious deceleration number of times, night trip durations account for total travel time ratio, 90km/h Above speed travel time accounting.

Premium based on mileage and driving behavior safety the most according to claim 1 reduces method, it is characterised in that institute Stating in step (1), described vehicle sensor data is by installing vehicle-mounted T-BOX equipment or installing OBD after vehicle before vehicle Equipment obtains.

Premium based on mileage and driving behavior safety the most according to claim 6 reduces method, it is characterised in that institute State vehicle sensor data and include Vehicle Identify Number, data receipt time, longitude, latitude, kilometrage, the tire pressure of front revolver, front right The tire pressure of wheel, the tire pressure of rear revolver, the tire pressure of rear right wheel, instantaneous oil consumption, longitudinal acceleration, with the position of steering wheel of corner metering Put, the rotating speed of steering wheel, speed, engine speed, and front left-hand door, front right door, rear left-hand door, rear right door, boot, steering wheel The state of sensor, described state is for opening or closing.