CN108268678A - Driving behavior analysis method, apparatus and system - Google Patents
Driving behavior analysis method, apparatus and system Download PDFInfo
- Publication number
- CN108268678A CN108268678A CN201611262696.0A CN201611262696A CN108268678A CN 108268678 A CN108268678 A CN 108268678A CN 201611262696 A CN201611262696 A CN 201611262696A CN 108268678 A CN108268678 A CN 108268678A
- Authority
- CN
- China
- Prior art keywords
- driving behavior
- behavior analysis
- vehicle
- analysis model
- collision
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/30—Circuit design
- G06F30/36—Circuit design at the analogue level
- G06F30/367—Design verification, e.g. using simulation, simulation program with integrated circuit emphasis [SPICE], direct methods or relaxation methods
Landscapes
- Engineering & Computer Science (AREA)
- Computer Hardware Design (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Microelectronics & Electronic Packaging (AREA)
- Evolutionary Computation (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Traffic Control Systems (AREA)
Abstract
A kind of driving behavior analysis method, apparatus and system.The method includes:Obtain the driving range of vehicle vehicle in the driving behavior data and first duration in default first duration;Driving behavior data in first duration are handled, obtain the number that each driving behavior occurs in default mileage;The number that obtained each driving behavior occurs in default mileage is input to the driving behavior analysis model trained, the collision probability of happening of the vehicle is calculated using the driving behavior analysis model and is exported, wherein, the input parameter of the driving behavior analysis model includes driving behavior, between the driving behavior and the collision probability of happening of the vehicle there are the first mapping relations, each driving behavior corresponding weight in first mapping relations in the driving behavior analysis model is obtained by training.Using the above scheme, the accuracy and traffic safety of driving behavior analysis can be improved.
Description
Technical field
The present invention relates to car networking technology field more particularly to a kind of driving behavior analysis method, apparatus and system.
Background technology
With social development, automobile has become people's essential vehicles in daily traffic.At present, more
The driving behavior analysis of vehicle can be used under kind situation.For example, insurance company is analyzed according to the driving behavior of vehicle, and
Vehicle premium is determined according to driving behavior analysis result.For another example, some enterprises are when recruiting driver, according to the driving behavior of driver
To determine the driving experience of driver.
However, the accuracy of driving behavior analysis result that existing driving behavior analysis method obtains is relatively low.
Invention content
Present invention solves the technical problem that it is how to improve the accuracy and traffic safety of driving behavior analysis.
Include in order to solve the above technical problems, the embodiment of the present invention provides a kind of driving behavior analysis method:Obtain vehicle
The driving range of the vehicle in driving behavior data and first duration in default first duration;To described first
Driving behavior data in duration are handled, and obtain the number that each driving behavior occurs in default mileage;The institute that will be obtained
It states the number that each driving behavior occurs in default mileage and is input to the driving behavior analysis model trained, using the driving
Analysis model of network behaviors calculates the collision probability of happening of the vehicle and exports, wherein, the input of the driving behavior analysis model
Parameter includes driving behavior, has the first mapping relations between the driving behavior and the collision probability of happening of the vehicle, leads to
It crosses training and obtains each driving behavior corresponding weight in first mapping relations in the driving behavior analysis model.
Optionally, it after the collision probability of happening for calculating the vehicle, further includes:Occurred according to the collision of the vehicle
Probability with reference to the driving behavior data in first duration, using the driving behavior analysis model, analyzes each driving behavior
Influence to the collision probability of happening of the vehicle;For influence of each driving behavior to vehicle collision probability of happening, generation is driven
It sails behavior Improving advice and exports.
Optionally, the driving behavior includes:Anxious acceleration, anxious deceleration, idling, fatigue driving and hypervelocity.
Optionally, the driving behavior data in first duration are respectively processed according to time interval, obtain phase
Answer the number that each driving behavior occurs in default mileage in time interval;By each driving in obtained corresponding time interval
The number that behavior occurs in default mileage is input to the driving behavior analysis model trained;Using the driving behavior analysis
Model calculates the collision probability of happening of the vehicle and exports, and the input parameter of the driving behavior analysis model is when different
Between section have corresponding weight.
Optionally, the input parameter of the driving behavior analysis model further includes collision accident relation factor, the vehicle
Collision probability of happening and the driving behavior and collision accident relation factor there are the second mapping relations, by training respectively
Into the driving behavior analysis model, each driving behavior and collision accident relation factor are corresponding in second mapping relations
Weight.
Optionally, the method further includes:Obtain the corresponding collision accident association of the vehicle in first duration because
Prime number evidence;Collision accident relation factor data in first duration are handled;By treated, collision accident is associated with
Factor data is input to the driving behavior analysis model trained.
Optionally, the collision accident relation factor includes following at least one:Zig zag, Frequent Accidents, road conditions, gas
Time, driving age, personality, region and record violating the regulations.
Optionally, the driving behavior analysis model is trained in the following way:Obtain training sample, the training sample
Include driving behavior data, collision accident relation factor data and the degree value to collide;The degree value of collision is more than
Or positive sample is marked equal to the training sample of predetermined threshold value, the training sample that the degree value of collision is less than to the threshold value is labeled as
Negative sample;Using logistic regression algorithm, according to the driving behavior data and collision accident relation factor data, to the positive sample
This and the negative sample carry out logistic regression training, obtain the driving behavior analysis model.
Optionally, the driving behavior analysis model is trained in the following way:Obtain training sample, the training sample
Include driving behavior data and the degree value to collide;The degree value of collision is greater than or equal to the training sample mark of threshold value
Remember positive sample, the training sample that the degree value of collision is less than to the threshold value is labeled as negative sample;Using logistic regression algorithm, root
According to the driving behavior data, logistic regression training is carried out to the positive sample and the negative sample, obtains the driving behavior
Analysis model.
Optionally, when to the driving behavior analysis model training, including:It is selected at random from the training sample pre-
If the training sample of number is as test sample;The driving behavior analysis model is carried out using test sample verification accurate
Exactness is verified.
It is optionally, described that accuracy validation is carried out to the driving behavior analysis model using test sample verification,
Including:The test sample is randomly divided into N groups test subsample, N is natural number, and N >=3;It is surveyed using wherein arbitrary N-1 groups
Swab sample carries out accuracy validation to one group of test subsample except the N-1 groups, until N groups test subsample is equal
It is verified.
Optionally, the collision probability of happening for calculating the vehicle, including:Obtain adjusting thresholds coefficient;According to described
Adjusting thresholds coefficient obtains each driving behavior in first mapping relations from the driving behavior analysis model trained
Weight corresponding to middle difference, wherein, it is each to drive under different adjusting thresholds coefficients in the driving behavior analysis model
Behavior corresponds to different weights respectively;According to each driving behavior got, institute is right in first mapping relations respectively
The weight answered calculates the collision probability of happening of the vehicle.
Optionally, the method further includes:When monitoring preset model modification trigger event, the vehicle is obtained
Driving behavior data are used as more new samples, and training is updated to the driving behavior analysis model using the more new samples,
And the driving behavior analysis model trained and obtained will be updated as the driving behavior analysis model trained.
Optionally, described first when it is a length of using current time as the period of end of time.
The embodiment of the present invention also provides a kind of driving behavior analysis device and includes:Acquiring unit, processing unit, input are single
Member, driving behavior analysis model, computing unit and the first output unit, wherein:The acquiring unit is preset suitable for obtaining vehicle
The driving range of the vehicle in driving behavior data and first duration in first duration;The processing unit is fitted
It is handled in the driving behavior data in first duration, obtains time that each driving behavior occurs in default mileage
Number;The input unit, the number suitable for obtained each driving behavior is occurred in default mileage, which is input to, has trained
The driving behavior analysis model;The input parameter of the driving behavior analysis model includes driving behavior, described to drive row
Between the collision probability of happening of vehicle there are the first mapping relations, obtained in the driving behavior analysis model by training
Each driving behavior corresponding weight in first mapping relations;The computing unit, suitable for using the driving behavior point
Model is analysed, calculates the collision probability of happening of the vehicle;First output unit, suitable for exporting the computing unit calculating
The collision probability of happening of the vehicle.
Optionally, described device further includes:Analytic unit and the second output unit, wherein:The analytic unit, suitable for
After the collision probability of happening for calculating the vehicle, according to the collision probability of happening of the vehicle, with reference in first duration
The corresponding data of all driving behaviors, using the driving behavior analysis model, analyze each driving behavior to the vehicle
Collision probability of happening influence, for influence of each driving behavior to vehicle collision probability of happening, generation driving behavior improves
It is recommended that;Second output unit, suitable for exporting the driving behavior Improving advice of the analytic unit generation.
Optionally, the processing unit, suitable for dividing according to time interval the driving behavior data in first duration
It is not handled, obtains the number that each driving behavior occurs in default mileage in corresponding time interval;The input unit is fitted
The number that each driving behavior occurs in default mileage in by obtained corresponding time interval is input to driving of having trained
Sail Analysis model of network behaviors;The computing unit suitable for using the driving behavior analysis model, calculates the collision hair of the vehicle
Raw probability, the input parameter of the driving behavior analysis model have corresponding weight in different time intervals.
Optionally, the input parameter of the driving behavior analysis model further includes collision accident relation factor, the vehicle
Collision probability of happening and the driving behavior and collision accident relation factor there are the second mapping relations, by training respectively
Into the driving behavior analysis model, each driving behavior and collision accident relation factor are corresponding in second mapping relations
Weight.
Optionally, the acquiring unit is further adapted for obtaining the corresponding collision accident of the vehicle in first duration and closes
Connection factor;The processing unit is further adapted for handling the collision accident relation factor data in first duration;It is described
Input unit is further adapted for treated driving behavior analysis model that collision accident relation factor data, which are input to, has trained.
Optionally, described device further includes:First training unit, suitable for training the driving behavior point in the following way
Analyse model:Training sample is obtained, the training sample includes driving behavior data, collision accident relation factor data and generation
The degree value of collision;The training sample that the degree value of collision is greater than or equal to predetermined threshold value marks positive sample, by the journey of collision
The training sample that angle value is less than the threshold value is labeled as negative sample;Using logistic regression algorithm, according to the driving behavior data
And collision accident relation factor data, logistic regression training is carried out to the positive sample and the negative sample, obtains the driving
Analysis model of network behaviors.
Optionally, described device further includes:Second training unit, suitable for training the driving behavior point in the following way
Analyse model:Training sample is obtained, the training sample includes driving behavior data and the degree value to collide;By collision
Degree value is greater than or equal to the training sample label positive sample of threshold value, and the degree value of collision is less than to the training sample of the threshold value
Labeled as negative sample;Using logistic regression algorithm, according to the driving behavior data, to the positive sample and the negative sample into
Row logistic regression is trained, and obtains the driving behavior analysis model.
Optionally, described device further includes test cell, suitable for selecting preset number at random from the training sample
Training sample is as test sample;Accuracy is carried out using test sample verification to the driving behavior analysis model to test
Card.
Optionally, the test cell, suitable for the test sample is randomly divided into N groups test subsample, N is nature
Number, and N >=3;It is accurate that one group of test subsample except the N-1 groups is carried out using wherein arbitrary N-1 groups test subsample
Degree verification, until N groups test subsample is verified.
Optionally, the computing unit, suitable for obtaining adjusting thresholds coefficient;According to the adjusting thresholds coefficient, from described
The weight corresponding to each driving behavior respectively is obtained in the driving behavior analysis model trained, wherein, in the driving behavior
In analysis model, under different adjusting thresholds coefficients, each driving behavior corresponds to different in first mapping relations respectively
Weight;According to each driving behavior got weight corresponding in first mapping relations respectively, described in calculating
The collision probability of happening of vehicle.
Optionally, described device further includes:Updating unit, suitable for when monitoring preset model modification trigger event,
The driving behavior data for obtaining the vehicle are used as more new samples, using the more new samples to the driving behavior analysis model
Training is updated, and the driving behavior analysis model trained and obtained will be updated as the driving behavior analysis model trained.
The embodiment of the present invention also provides a kind of driving behavior analysis system, including:Data acquisition device and any of the above are driven
Behavioural analysis device is sailed, wherein:The data acquisition device, suitable for obtaining driving behavior number of the vehicle in default first duration
The driving range of the vehicle according to this and in first duration.
Optionally, the data acquisition device acquires institute suitable for passing through the driving states sensor installed on the vehicle
It states driving behavior data of the vehicle in default first duration or the vehicle is obtained from vehicle storage device default first
Driving behavior data in duration.
Compared with prior art, the technical solution of the embodiment of the present invention has the advantages that:
Using the driving behavior analysis model trained, using the number that each driving behavior occurs in default mileage as input
Parameter, according to each driving behavior corresponding weight in the first mapping relations in the driving behavior analysis model, described in calculating
The collision probability of happening of vehicle.In the average extremely default mileage of number that driving behavior each in first duration is occurred, so as to
Influence of the single driving behavior to the collision probability being calculated can be reduced, so as to improve the accurate of driving behavior analysis
Property, improve traffic safety.
Further, by the collision probability of happening of the vehicle, with reference to the driving behavior data in first duration,
Influence of each driving behavior to the collision probability of happening of the vehicle is analyzed, obtains bad steering behavior, and provide driving behavior
Improving advice can contribute to remind and improve the bad steering behavior of user, reduce the collision probability of happening of vehicle, improve row
Vehicle safety.
Further, the driving behavior data in first duration are respectively processed according to time interval, obtained
The number that each driving behavior occurs in default mileage in corresponding time interval, so as to respect to each driving behavior in difference
Time interval is different to the influence degree to collide, further improves the accuracy of driving behavior analysis.
Further, the input parameter in the driving behavior analysis model can also include the vehicle in the first duration
Corresponding collision accident relation factor data, then when calculating the collision probability of vehicle, comprehensive collision accident relation factor can be into
One step improves the accuracy of driving behavior analysis.
Further, by selected threshold regulation coefficient, each drive to select in the driving behavior analysis model is gone
For corresponding weight, so as to meet different user under different application scene to precision the needs of.
Further, when monitoring preset model modification trigger event, the driving behavior data of the vehicle are obtained
As more new samples, it is updated training to the driving behavior analysis model, and by updated driving behavior analysis model
As the driving behavior analysis model trained, periodically the driving behavior analysis model is updated, can be caused described
The matching degree higher of driving behavior analysis model and the vehicle, so as to further improve the accurate of driving behavior analysis
Degree.
Description of the drawings
Fig. 1 is a kind of flow chart of driving behavior analysis method in the embodiment of the present invention;
Fig. 2 is a kind of structure diagram of driving behavior analysis device in the embodiment of the present invention;
Fig. 3 is the structure diagram of another driving behavior analysis device in the embodiment of the present invention.
Specific embodiment
The usually real-time collection vehicle data of existing driving behavior analysis method, and according to real-time collected vehicle number
According to, the driving behavior of user is analyzed, since under different scenes, the influence factor of the driving behavior of user is different,
Obtained driving behavior cannot react the practical driving situation of user, and accuracy is relatively low.
To solve the above problems, in embodiments of the present invention, the driving behavior analysis model that use has been trained, with each driving
The number that behavior occurs in default mileage is input parameter, according to each driving behavior in the driving behavior analysis model the
Corresponding weight in one mapping relations calculates the collision probability of happening of the vehicle.By each driving behavior in first duration
The number of generation is average to presetting in mileage, so as to reduce shadow of the single driving behavior to the collision probability being calculated
It rings, so as to improve the accuracy of driving behavior analysis, improves traffic safety.
It is understandable for above-mentioned purpose, feature and advantageous effect of the invention is enable to become apparent, below in conjunction with the accompanying drawings to this
The specific embodiment of invention is described in detail.
With reference to Fig. 1, a kind of flow chart of driving behavior analysis method in the embodiment of the present invention is given, with reference to specific
Step is described in detail.
Step 11, vehicle vehicle in driving behavior data and first duration preset in the first duration is obtained
Driving range.
In specific implementation, it can be obtained by the vehicle diagnosing system (On-Board Diagnostic, OBD) of vehicle
Driving behavior data and driving range of the vehicle in first duration, can also pass through vehicle-mounted TBOX and the phase in vehicle
Controller is answered to communicate, obtains driving behavior data and driving range of the vehicle in first duration, it can be with
By the various sensors installed on vehicle driving behavior data are obtained in real time, and store.It is understood that it can also pass through
Other Intelligent hardwares installed on vehicle obtain driving behavior data and driving range of the vehicle in first duration.
In an embodiment of the present invention, the driving behavior can include:It is anxious accelerate, anxious deceleration, idling, fatigue driving,
Hypervelocity etc..In an alternative embodiment of the invention, the driving behavior includes:Neutral position sliding, engine prewarming overlong time, it is long when
Between cannot not stop flame-outly, running-in period travel speed is too fast or parking when keep D grade step on brake etc..It in specific implementation, can basis
Big data counts the influence of various driving behaviors to the collision probability of happening of vehicle, and using the higher driving behavior of the degree of association as
The input parameter of driving behavior analysis model used by during driving behavior analysis, and it is not limited to enumerated above drive
Sail behavior.
In specific implementation, first duration can be nature week, calendar month, nature season, can also be as needed
Specific duration is set.For example, 45 days, 52 days etc., the specific value of first duration is not limited in embodiments of the present invention
It is fixed.
In specific implementation, the beginning and end of first duration can also be set according to actual needs.When need
When checking the driving behavior in certain time period, can be set according to the period to be viewed first duration starting point and
Terminal.In an embodiment of the present invention, described first when it is a length of using current time as the period of end of time.For example, first
Shi Changwei mono- month, current time are 22 days 11 December in 2016:25:36, then first duration is from 2016 11
On the moon 23 11:25:00~2016 on December 22,11:24:59.
Step 12, the driving behavior data in first duration are handled, obtains each driving behavior in default
The number occurred in journey.
In specific implementation, the driving behavior data in first duration that gets can be handled, calculated
The total degree that each driving behavior occurs in the first duration, then according to the driving range in first duration, calculating is respectively driven
Sail the number that behavior occurs in default mileage.
For example, occurring suddenly to accelerate m times in a calendar month, anxious deceleration n times, the driving range in a calendar month is
P kilometers, then anxious accelerate occurs in default Q kilometers of mileageIt is secondary, anxious slow down occursIt is secondary, wherein, m, n, P, Q are equal
For positive integer, and P >=Q.
Step 13, the number that each driving behavior being calculated occurs in default mileage is input to what is trained
Driving behavior analysis model calculates the collision probability of happening of the vehicle using the driving behavior analysis model and exports.
In specific implementation, the input parameter of the driving behavior analysis model can include driving behavior, the driving
There are the first mapping relations between behavior and the collision probability of happening of the vehicle.The driving behavior can be obtained by training
Each driving behavior corresponding weight in first mapping relations in analysis model, the weight are used to characterize each driving behavior
The degree of correlation between the collision probability of happening of vehicle.
The number that obtained each driving behavior is occurred in default mileage is input to and has trained as input parameter
Driving behavior analysis model in, it is corresponding in first mapping relations according to driving behavior each in driving behavior analysis model
Weight, calculate the collision probability of happening of the vehicle.After the collision probability of happening that the vehicle is calculated, tied calculating
Fruit exports.
As shown in the above, the driving behavior analysis model that use has been trained, with each driving behavior in default mileage
The number of generation is input parameter, corresponding in the first mapping relations according to each driving behavior in the driving behavior analysis model
Weight, calculate the collision probability of happening of the vehicle.The number that driving behavior each in first duration is occurred is average extremely
In default mileage, so as to reduce influence of the single driving behavior to the collision probability being calculated, driven so as to improve
The accuracy of behavioural analysis is sailed, improves traffic safety.
In specific implementation, it is real in we's invention one in order to more intuitively show the analysis result of driving behavior to user
It applies in example, vehicle can be set to collide the correspondence between probability and driving behavior assessed value.Vehicle is being calculated
Collision probability of happening after, can the collision probability of happening of vehicle be converted by corresponding assessment according to preset correspondence
Value.
In an embodiment of the present invention, driving behavior assessed value may be used 0~100 numerical value and be indicated, and drive row
It is the collision probability of happening of assessed value and vehicle into inverse correlation.Driving behavior assessed value is bigger, shows that driving behavior is better, described
The collision probability of happening of vehicle is smaller, conversely, driving behavior assessed value is smaller, shows that bad steering behavior is more, the vehicle
Collision probability of happening it is bigger.
In specific implementation, in order to improve the bad steering behavior of user, reduce the collision probability of happening of vehicle and improve
Traffic safety, in an embodiment of the present invention, can also be according to the vehicle after the collision probability of happening of the vehicle is obtained
Collision probability of happening, with reference to the driving behavior data in first duration, using the driving behavior analysis model, point
Influence of each driving behavior to the collision probability of happening of the vehicle is analysed, for each driving behavior to vehicle collision probability of happening
It influences, generate driving behavior Improving advice and exports.
For example, being shown in driving behavior data in the first duration, the number suddenly to slow down is 20 times, in the vehicle
It collides in probability of happening, anxious deceleration is larger to the impact probability to collide, therefore can provide driving behavior Improving advice:As possible
Reduce the number suddenly to slow down.In specific implementation, can driving behavior Improving advice be exported by associated car entertainment device,
Can also by the mobile terminals such as mobile phone export driving behavior Improving advice, can also be exported by other-end equipment or
Driving behavior Improving advice is exported with audible by audio devices.
In practical applications, since in section in different times, identical driving behavior is to the collision probability of happening of vehicle
It is generated to influence difference.For example, influence of the fatigue driving to the collision probability of happening of vehicle is more than daytime fatigue at night
It drives.For another example, the anxious urgency for accelerating to occur when being more than night to the collision probability of happening of vehicle occurred when going to work peak period adds
Speed.Therefore, if in the collision probability of happening for calculating vehicle, no matter within any time, the corresponding power of same driving behavior
Weight all same may influence the accuracy of the collision probability of happening for the vehicle being calculated namely influence to the vehicle
The accuracy of driving behavior analysis.
In order to the vehicle improved collide probability accuracy and improve to the driving behavior analysis of vehicle
Accuracy.In an embodiment of the present invention, the driving behavior data in first duration are carried out respectively according to time interval
Processing obtains the number that each driving behavior occurs in default mileage in corresponding time interval, the corresponding time interval that will be obtained
The number that interior each driving behavior occurs in default mileage is input to the driving behavior analysis model trained;Using described
Driving behavior analysis model calculates the collision probability of happening of the vehicle and exports.The input of the driving behavior analysis model
Parameter has corresponding weight in different time intervals.
In specific implementation, setting duration can be divided, obtained different according to preset time dividing mode
Time interval segments each driving behavior respectively to the number occurred in each time interval.According to each driving behavior when each
Between the driving range of the number and the vehicle that are occurred in section in first duration, obtain in the corresponding time
The number that each driving behavior occurs in default mileage in section.In the process being trained to the driving behavior analysis model
In, corresponding weight may be different in different time intervals for same driving behavior.
For example, each consecutive days are divided into the morning, afternoon and evening on time dimension.Wherein, when the morning is corresponding
Between section be 05:00:00~11:59:59, afternoon, corresponding time interval was 12:00:00~18:59:59, it is corresponding at night
Time interval is 17:00:00~24:00:00 and 00:00:00~04:59:59.Weight of the fatigue driving in the morning, afternoon
Weight and the weight in evening can be different, and weight at night be respectively greater than in the weight in the morning and the power in afternoon
Weight.Include the time data of each driving behavior generation, the time number occurred according to each driving behavior in each driving behavior data
According to the number occurred in the corresponding time interval of each driving behavior can be obtained.
In specific implementation, since the environment in vehicle travel process is complex, the collision probability of happening of vehicle is not
It is only influenced, can also be influenced by other various factors by driving behavior.To cause the analysis to driving behavior more
Add accurately, the probability that collides of obtained vehicle is more accurate.In an embodiment of the present invention, the driving behavior analysis mould
The input parameter of type can also include collision accident relation factor, the collision probability of happening of the vehicle and the driving behavior and
Collision accident relation factor has the second mapping relations, is respectively obtained in the driving behavior analysis model by training and respectively driven
Behavior and collision accident the relation factor corresponding weight in second mapping relations.
In specific implementation, the corresponding collision accident relation factor number of the vehicle in first duration can be obtained
According to handling the collision accident relation factor data in first duration;It will treated collision accident relation factor
Data are input to the driving behavior analysis model trained.
In specific implementation, collision accident relation factor data of the vehicle in first duration can be obtained.
The relationship between each collision accident relation factor and the collision probability of happening of vehicle can be determined by big data statistical analysis,
Correlation analysis can also be used, to determine the pass between each collision accident relation factor and the collision probability of happening of vehicle
System, and determine which factor may have an impact the collision probability of happening of vehicle, it can also determine each collision time association
Factor is to the influence degree of the collision probability of happening of vehicle.
In specific implementation, the collision accident relation factor can include zig zag, Frequent Accidents, road conditions, weather,
It is one or more in driving age, personality, region and record violating the regulations.For example, road conditions whether congestion and road conditions it is whether smooth,
Fresh driver still has the driver of certain driving experience during driver, and driver is northerner or southerner, by personality
It influences, southerner drives and violating the regulations record etc. such as make a dash across the red light, drive in the wrong direction more careful than northerner.
In specific implementation, a large amount of training sample may be used to be trained driving behavior analysis model, obtain each
The weight of driving behavior and each collision accident relation factor corresponding in second mapping relations.
In specific implementation, various ways may be used to be trained the driving behavior analysis model.
In an embodiment of the present invention, training sample is obtained, the training sample includes driving behavior data, collision thing
Part relation factor data and the degree value to collide.The degree value of collision is greater than or equal to the training sample mark of predetermined threshold value
Remember positive sample, the training sample that the degree value of collision is less than to the threshold value is labeled as negative sample.Using logistic regression algorithm, root
According to the driving behavior data and collision accident relation factor data, logistic regression is carried out to the positive sample and the negative sample
Training, obtains the driving behavior analysis model.
In specific implementation, 0~100 numeric form expression can be used in the degree value of collision, can also be using 0~1
Numeric form represents, it is to be understood that the degree value of collision can also have other representations, specifically can be as needed
It is set, is not limited herein.Predetermined threshold value can be set according to the representation of the degree value of collision.For example, it touches
When the numeric form that the degree value hit can be used 0~100 represents, predetermined threshold value can be taken as 26.It is understood that default threshold
The specific value of value according to practical application scene and can be set, or the other values such as 30,40, herein to tool
Body value does not limit.
In an alternative embodiment of the invention, training sample is obtained, the training sample includes driving behavior data and hair
The degree value of raw collision, the training sample that the degree value of collision is greater than or equal to threshold value marks positive sample, by the degree of collision
Value is labeled as negative sample less than the training sample of the threshold value.It is right according to the driving behavior data using logistic regression algorithm
The positive sample and the negative sample carry out logistic regression training, obtain the driving behavior analysis model.
In an embodiment of the present invention, the driving behavior analysis mould is obtained using the training of following logistic regression formula (1)
Type.
Wherein, g (X) represents the collision probability of happening of vehicle;xiFor input parameter;θiFor input parameter xiCorresponding weight;
θ0For constant;1≤i≤n, i and n are positive integer.
In specific implementation, during driving behavior analysis model training, in order to improve the accurate of logistic regression type
Degree, to improve the accuracy of driving behavior analysis, in an embodiment of the present invention, in driving behavior analysis model training process
In, the training sample of preset number can be selected at random from the training sample as test sample, using the test specimens
This verification carries out accuracy validation to the driving behavior analysis model.
In specific implementation a certain proportion of training sample can be chosen as test sample from the training sample.
For example, 20% is chosen from training sample is used as test sample.
In specific implementation, following manner may be used, accuracy validation is carried out to the driving behavior analysis model.It will
The test sample is randomly divided into N groups test subsample, and N is natural number, and N >=3.Using wherein arbitrary N-1 groups test increment
This carries out accuracy validation to one group of test subsample except the N-1 groups, until N groups test subsample is tested
Card.
In an embodiment of the present invention, the test sample is randomly divided into three groups of test subsamples, using two groups of tests
Subsample carries out accuracy validation to another set test subsample, until three groups of test subsamples are verified.
For example, the test sample is divided into tri- groups of A, B and C, two groups C groups are verified, examined using institute using A, B
State driving behavior analysis model in C groups test sample estimate collision probability compared with the actual collision probability of C groups whether
Meet accuracy requirement.Equally, using A, C, two groups are verified B groups, two groups A groups are verified using B, C.
In specific implementation, different application scenarios, different users to the precise requirements of driving behavior analysis not yet
Together, in order to meet the required precision of several scenes, in an embodiment of the present invention, in the collision probability of happening for calculating the vehicle
When, adjusting thresholds coefficient can be obtained, according to the adjusting thresholds coefficient, from the driving behavior analysis model trained
Each driving behavior weight corresponding in first mapping relations respectively is obtained, wherein, in the driving behavior analysis mould
In type, under different adjusting thresholds coefficients, driving behavior corresponds to different weights respectively.Each drive according to getting is gone
For weight corresponding in first mapping relations respectively, the collision probability of happening of the vehicle is calculated.
In specific implementation, under different adjusting thresholds coefficients, each driving behavior is right in first mapping relations
The weight answered also differs.In different adjusting thresholds coefficients, corresponding weight identical also can may be used for same driving behavior
With difference.The accuracy of collision probability of happening being calculated using different adjusting thresholds coefficients is different.
For example, adjusting thresholds coefficient is Z1, Z2 and Z3, and Z1 ≠ Z2 ≠ Z3, and when adjusting thresholds coefficient is Z1, anxious acceleration pair
The weight answered is a1, and anxious corresponding weight of slowing down is b1;It is anxious to accelerate corresponding weight as a2 when adjusting thresholds coefficient is Z2, it is anxious
Corresponding weight of slowing down is b2;Anxious to accelerate corresponding weight as a3 when adjusting thresholds coefficient is Z3, anxious corresponding weight of slowing down is
b3.In specific implementation, a1, a2 and a3 can be differed, and b1, b2 and b3 can also be differed.Or in a1, a2 and a3
Any two of which can be identical, and b1, b2 and b3 can be differed.Or any two of which can phase in b1, b2 and b3
Together, a1, a2 and a3 can be differed.Or in a1, a2 and a3 three can be identical, b1, b2 and b3 can be differed.
In practical applications, there may also be other situations, only need to be under different adjusting thresholds coefficients, each driving behavior
Corresponding weight is not exactly the same.
In specific implementation, when monitoring preset model modification trigger event, the driving behavior of the vehicle is obtained
Data are used as more new samples, training are updated to the driving behavior analysis model, and will update using the more new samples
The driving behavior analysis model that training obtains is as the driving behavior analysis model trained.By timing to the driving behavior
Analysis model is updated, and can cause the matching degree higher of the driving behavior analysis model and the vehicle, so as to
Further improve the accuracy of driving behavior analysis.
In specific implementation, preset model modification trigger event can be default second duration, whenever reaching second
When long, the driving behavior analysis model is updated automatically.Second duration can be two months, or one
Season can also be 1 year.The setting of the specific value of second duration can be according to the value of first duration, Huo Zhegen
Factually the application scenarios on border are set.
Preset model modification trigger condition may be the situation of change in region residing for vehicle, for example, the previous existence of user
It is living to be settled down later to northeast in south, due to northeast environment and it is southern differ greatly, can be according to the vehicle master detected
The zone of action wanted is updated driving behavior analysis model.Preset model modification trigger condition may be user according to
Own situation is independently updated.
For those skilled in the art is caused to be better understood from and realize the present invention, this law inventive embodiments additionally provide one kind
Driving behavior analysis device.
With reference to Fig. 2, a kind of structure diagram of driving behavior analysis device in the embodiment of the present invention is given.The driving
Behavioural analysis device 20 can include:Acquiring unit 21, processing unit 22, input unit 23, driving behavior analysis model 24, meter
25 and first output unit 26 of unit is calculated, wherein:
The acquiring unit 21, suitable for obtain vehicle preset driving behavior data in the first duration and it is described first when
The driving range of the vehicle in length;
The processing unit 22 suitable for handling the driving behavior data in first duration, obtains each driving
The number that behavior occurs in default mileage;
The input unit 23, the number suitable for obtained each driving behavior is occurred in default mileage are input to
The driving behavior analysis model trained;
The input parameter of the driving behavior analysis model 24 includes driving behavior, the driving behavior and the collision of vehicle
There are the first mapping relations, obtaining each driving behavior in the driving behavior analysis model 24 by training exists between probability of happening
Corresponding weight in first mapping relations;
The computing unit 25, suitable for using the driving behavior analysis model 24, the collision for calculating the vehicle occurs
Probability;
First output unit 26 occurs generally suitable for exporting the collision for the vehicle that the computing unit 25 calculates
Rate.
In specific implementation, the driving behavior can include:Anxious acceleration, anxious deceleration, idling, fatigue driving and hypervelocity.
From the foregoing, it will be observed that using the driving behavior analysis model trained, occurred in default mileage with each driving behavior
Number is input parameter, according to each driving behavior corresponding power in the first mapping relations in the driving behavior analysis model
Weight calculates the collision probability of happening of the vehicle.The number that driving behavior each in first duration is occurred is average to default
In mileage, so as to reduce influence of the single driving behavior to the collision probability being calculated, row is driven so as to improve
For the accuracy of analysis, traffic safety is improved.
The structure diagram of another driving behavior analysis device in the embodiment of the present invention provided with reference to Fig. 3.Specific real
Shi Zhong, the driving behavior analysis device 20 can also include on the basis of Fig. 2:31 and second output unit of analytic unit
32, wherein:
The analytic unit 31, suitable for after the collision probability of happening for calculating the vehicle, according to touching for the vehicle
Probability of happening is hit, with reference to the corresponding data of driving behaviors all in first duration, using the driving behavior analysis
Model 24 analyzes influence of each driving behavior to the collision probability of happening of the vehicle, for each driving behavior to vehicle collision
The influence of probability of happening generates driving behavior Improving advice;
Second output unit 32, suitable for exporting the driving behavior Improving advice of the analytic unit generation.
In specific implementation, the processing unit 22, suitable for according to time interval to the driving row in first duration
It is respectively processed for data, obtains the number that each driving behavior occurs in default mileage in corresponding time interval;
The input unit 23, suitable for each driving behavior in obtained corresponding time interval is sent out in default mileage
Raw number is input to the driving behavior analysis model 24 trained;
The computing unit 25, suitable for using the driving behavior analysis model 24, the collision for calculating the vehicle occurs
Probability, the input parameter of the driving behavior analysis model 24 have corresponding weight in different time intervals.
In specific implementation, the input parameter of the driving behavior analysis model 24 can include collision accident association because
Element, collision probability of happening and the driving behavior and collision accident relation factor of the vehicle have the second mapping relations, lead to
It crosses training and respectively obtains in the driving behavior analysis model 24 each driving behavior and collision accident relation factor described second
Corresponding weight in mapping relations.
In specific implementation, the acquiring unit 21, is further adapted for obtaining in first duration that the vehicle is corresponding to be touched
Hit event correlation factor;
The input unit 23 is further adapted for treated driving that collision accident relation factor data, which are input to, has trained
Analysis model of network behaviors 24, the input parameter of the driving behavior analysis model include collision accident relation factor;The processing is single
Member 22, is further adapted for handling the collision accident relation factor data in first duration;The input unit 23, it is also suitable
In by treated driving behavior analysis model 24 that collision accident relation factor data, which are input to, has trained.
In specific implementation, the collision accident relation factor can include zig zag, Frequent Accidents, road conditions, weather,
It is one or more in driving age, personality, region and record violating the regulations.
In specific implementation, the driving behavior analysis device 20 can also include the first training unit 33.Described first
Training unit 33, suitable for training the driving behavior analysis model 24 in the following way:Obtain training sample, the trained sample
This includes driving behavior data, collision accident relation factor data and the degree value to collide;The degree value of collision is big
In or equal to predetermined threshold value training sample label positive sample, by the degree value of collision be less than the threshold value training sample mark
For negative sample;Using logistic regression algorithm, according to the driving behavior data and collision accident relation factor data, to it is described just
Sample and the negative sample carry out logistic regression training, obtain the driving behavior analysis model 24.
In an embodiment of the present invention, logistic regression formula (1) instruction provided in the above embodiment of the present invention is provided
Get the driving behavior analysis model 24.
In specific implementation, the driving behavior analysis device 20 can also include the second training unit 34.Described second
Training unit 34, suitable for training the driving behavior analysis model 24 in the following way:Obtain training sample, the trained sample
This includes driving behavior data and the degree value to collide;The degree value of collision is greater than or equal to the training sample of threshold value
Positive sample is marked, the training sample that the degree value of collision is less than to the threshold value is labeled as negative sample;Using the logistic regression
According to the driving behavior data, logistic regression training is carried out to the positive sample and the negative sample, obtains described drive for algorithm
Sail Analysis model of network behaviors 24.
In specific implementation, the driving behavior analysis device 20 can also include test cell 35, suitable for from the instruction
Practice in sample and select the training sample of preset number at random as test sample;It is verified using the test sample to the driving
Analysis model of network behaviors 24 carries out accuracy validation.
In specific implementation, the test cell 35, suitable for the test sample is randomly divided into N groups test subsample, N
For natural number, and N >=3;Using wherein arbitrary N-1 groups test subsample to one group of test subsample except the N-1 groups into
Row accuracy validation, until N groups test subsample is verified.
In specific implementation, the computing unit 25, suitable for obtaining adjusting thresholds coefficient;According to the adjusting thresholds system
Number obtains the weight corresponding to each driving behavior respectively from the driving behavior analysis model 24 trained, wherein, in institute
It states in driving behavior analysis model 24, under different adjusting thresholds coefficients, each driving behavior is respectively in first mapping relations
The different weight of middle correspondence;According to each driving behavior got power corresponding in first mapping relations respectively
Weight calculates the collision probability of happening of the vehicle.
In specific implementation, the driving behavior analysis device 20 can also include updating unit 36, suitable for monitoring
During preset model modification trigger event, the driving behavior data for obtaining the vehicle are used as more new samples, using the update
Sample is updated the driving behavior analysis model 24 training, and obtained driving behavior analysis model is trained to make by updating
For the driving behavior analysis model 24 trained.
In specific implementation, the driving behavior analysis device 20 can be an independent hardware device, or
Server can also be application software or client, in corresponding equipment.It is understood that described drive row
For analytical equipment 20, there may also be other forms.
In specific implementation, the concrete operating principle of the driving behavior analysis device 20 and workflow may refer to this
Description in the driving behavior analysis method provided in invention above-described embodiment, details are not described herein again.
The embodiment of the present invention also provides a kind of driving behavior analysis system, and the driving behavior analysis system can include:
Data acquisition device and any driving behavior analysis device as described in the above embodiment of the present invention.
In specific implementation, the data acquisition device can obtain driving behavior number of the vehicle in default first duration
The driving range of the vehicle according to this and in first duration.
In specific implementation, the data acquisition device, can be by the driving states sensor installed on the vehicle
Driving behavior data of the vehicle in default first duration are acquired, the vehicle can also be obtained from vehicle storage device and is existed
Driving behavior data in default first duration.
One of ordinary skill in the art will appreciate that all or part of step in the various methods of above-described embodiment is can
It is completed with instructing relevant hardware by program, which can be stored in a computer readable storage medium, storage
Medium can include:ROM, RAM, disk or CD etc..
Although present disclosure is as above, present invention is not limited to this.Any those skilled in the art are not departing from this
It in the spirit and scope of invention, can make various changes or modifications, therefore protection scope of the present invention should be with claim institute
Subject to the range of restriction.
Claims (27)
- A kind of 1. driving behavior analysis method, which is characterized in that including:Acquisition vehicle is in the driving behavior data and first duration in default first duration in the driving of the vehicle Journey;Driving behavior data in first duration are handled, obtain time that each driving behavior occurs in default mileage Number;The number that obtained each driving behavior occurs in default mileage is input to the driving behavior analysis mould trained Type calculates the collision probability of happening of the vehicle using the driving behavior analysis model and exports, wherein, the driving behavior The input parameter of analysis model includes driving behavior, has the between the driving behavior and the collision probability of happening of the vehicle It is right in first mapping relations to obtain each driving behavior in the driving behavior analysis model by training for one mapping relations The weight answered.
- 2. driving behavior analysis method according to claim 1, which is characterized in that occur in the collision for calculating the vehicle After probability, further include:According to the collision probability of happening of the vehicle, with reference to the driving behavior data in first duration, using the driving Analysis model of network behaviors analyzes influence of each driving behavior to the collision probability of happening of the vehicle;For influence of each driving behavior to vehicle collision probability of happening, generate driving behavior Improving advice and export.
- 3. driving behavior analysis method according to claim 1, which is characterized in that the driving behavior includes:Anxious acceleration, Anxious deceleration, idling, fatigue driving and hypervelocity.
- 4. driving behavior analysis method according to claim 3, which is characterized in that during according to time interval to described first Driving behavior data in length are respectively processed, and obtain what each driving behavior in corresponding time interval occurred in default mileage Number;The number that each driving behavior in obtained corresponding time interval occurs in default mileage is input to what is trained Driving behavior analysis model;Using the driving behavior analysis model, calculate the collision probability of happening of the vehicle and export, the driving behavior point The input parameter for analysing model has corresponding weight in different time intervals.
- 5. driving behavior analysis method according to claim 1, which is characterized in that the driving behavior analysis model it is defeated Enter parameter and further include collision accident relation factor, collision probability of happening and the driving behavior and collision accident of the vehicle are closed Connection factor has the second mapping relations, and each driving behavior and collision in the driving behavior analysis model are respectively obtained by training Event correlation factor corresponding weight in second mapping relations.
- 6. driving behavior analysis method according to claim 5, which is characterized in that further include:Obtain the corresponding collision accident relation factor data of the vehicle in first duration;Collision accident relation factor data in first duration are handled;By treated driving behavior analysis model that collision accident relation factor data, which are input to, has trained.
- 7. driving behavior analysis method according to claim 5, which is characterized in that the collision accident relation factor includes Following at least one:Zig zag, Frequent Accidents, road conditions, weather, the driving age, personality, region and record violating the regulations.
- 8. driving behavior analysis method according to claim 5, which is characterized in that train the driving in the following way Analysis model of network behaviors:Training sample is obtained, the training sample includes driving behavior data, collision accident relation factor data and touches The degree value hit;The training sample that the degree value of collision is greater than or equal to predetermined threshold value marks positive sample, and the degree value of collision is less than institute The training sample for stating threshold value is labeled as negative sample;Using logistic regression algorithm, according to the driving behavior data and collision accident relation factor data, to the positive sample And the negative sample carries out logistic regression training, obtains the driving behavior analysis model.
- 9. driving behavior analysis method according to claim 1, which is characterized in that train the driving in the following way Analysis model of network behaviors:Training sample is obtained, the training sample includes driving behavior data and the degree value to collide;The training sample that the degree value of collision is greater than or equal to threshold value marks positive sample, and the degree value of collision is less than the threshold The training sample of value is labeled as negative sample;Using logistic regression algorithm, according to the driving behavior data, logic is carried out to the positive sample and the negative sample and is returned Return training, obtain the driving behavior analysis model.
- 10. driving behavior analysis method according to claim 8 or claim 9, which is characterized in that the driving behavior analysis During model training, including:The training sample of preset number is selected at random from the training sample as test sample;Accuracy validation is carried out to the driving behavior analysis model using test sample verification.
- 11. driving behavior analysis method according to claim 10, which is characterized in that described to be tested using the test sample Card carries out accuracy validation to the driving behavior analysis model, including:The test sample is randomly divided into N groups test subsample, N is natural number, and N >=3;Accuracy validation is carried out to one group of test subsample except the N-1 groups using wherein arbitrary N-1 groups test subsample, Until N groups test subsample is verified.
- 12. driving behavior analysis method according to claim 1, which is characterized in that the collision for calculating the vehicle Probability of happening, including:Obtain adjusting thresholds coefficient;According to the adjusting thresholds coefficient, each driving behavior is obtained described from the driving behavior analysis model trained Weight corresponding respectively in first mapping relations, wherein, in the driving behavior analysis model, different adjusting thresholds systems Under several, each driving behavior corresponds to different weights respectively;According to each driving behavior got weight corresponding in first mapping relations respectively, the vehicle is calculated Collision probability of happening.
- 13. driving behavior analysis method according to claim 1, which is characterized in that further include:When monitoring preset model modification trigger event, the driving behavior data for obtaining the vehicle are used as more new samples, Training is updated to the driving behavior analysis model, and the driving behavior trained and obtained will be updated using the more new samples Analysis model is as the driving behavior analysis model trained.
- 14. driving behavior analysis method according to claim 1, which is characterized in that when described first it is a length of with it is current when Carve the period for end of time.
- 15. a kind of driving behavior analysis device, which is characterized in that including:Acquiring unit, input unit, drives row at processing unit For analysis model, computing unit and the first output unit, wherein:The acquiring unit presets institute in driving behavior data and first duration in the first duration suitable for acquisition vehicle State the driving range of vehicle;The processing unit suitable for handling the driving behavior data in first duration, obtains each driving behavior and exists The number occurred in default mileage;The input unit, the number suitable for obtained each driving behavior is occurred in default mileage, which is input to, has trained The driving behavior analysis model;The input parameter of the driving behavior analysis model includes driving behavior, and the driving behavior and the collision of vehicle occur general Between rate there are the first mapping relations, each driving behavior is obtained in the driving behavior analysis model described first by training Corresponding weight in mapping relations;The computing unit suitable for using the driving behavior analysis model, calculates the collision probability of happening of the vehicle;First output unit, suitable for exporting the collision probability of happening for the vehicle that the computing unit calculates.
- 16. driving behavior analysis device according to claim 15, which is characterized in that further include:Analytic unit and second Output unit, wherein:The analytic unit, suitable for after the collision probability of happening for calculating the vehicle, being occurred according to the collision of the vehicle Probability with reference to the corresponding data of driving behaviors all in first duration, using the driving behavior analysis model, divides Influence of each driving behavior to the collision probability of happening of the vehicle is analysed, for each driving behavior to vehicle collision probability of happening It influences, generates driving behavior Improving advice;Second output unit, suitable for exporting the driving behavior Improving advice of the analytic unit generation.
- 17. driving behavior analysis device according to claim 16, which is characterized in that the processing unit, suitable for according to Time interval is respectively processed the driving behavior data in first duration, obtains respectively driving row in corresponding time interval For the number occurred in default mileage;The input unit, suitable for time that each driving behavior in obtained corresponding time interval occurs in default mileage Number is input to the driving behavior analysis model trained;The computing unit suitable for using the driving behavior analysis model, calculates the collision probability of happening of the vehicle, described The input parameter of driving behavior analysis model has corresponding weight in different time intervals.
- 18. driving behavior analysis device according to claim 15, which is characterized in that the driving behavior analysis model Input parameter further includes collision accident relation factor, collision probability of happening and the driving behavior and the collision accident of the vehicle Relation factor has the second mapping relations, respectively obtains each driving behavior in the driving behavior analysis model by training and touches Hit event correlation factor corresponding weight in second mapping relations.
- 19. driving behavior analysis device according to claim 18, which is characterized in that the acquiring unit is further adapted for obtaining Take the corresponding collision accident relation factor of the vehicle in first duration;The processing unit is further adapted for handling the collision accident relation factor data in first duration;The input unit is further adapted for treated driving behavior point that collision accident relation factor data, which are input to, has trained Analyse model.
- 20. driving behavior analysis device according to claim 18, which is characterized in that further include:First training unit is fitted In the trained driving behavior analysis model in the following way:Training sample is obtained, the training sample includes driving behavior data, collision accident relation factor data and touches The degree value hit;The training sample that the degree value of collision is greater than or equal to predetermined threshold value marks positive sample, and the degree value of collision is less than institute The training sample for stating threshold value is labeled as negative sample;Using logistic regression algorithm, according to the driving behavior data and collision accident relation factor data, to the positive sample And the negative sample carries out logistic regression training, obtains the driving behavior analysis model.
- 21. driving behavior analysis device according to claim 15, which is characterized in that further include:Second training unit is fitted In the trained driving behavior analysis model in the following way:Training sample is obtained, the training sample includes driving behavior data and the degree value to collide;The training sample that the degree value of collision is greater than or equal to threshold value marks positive sample, and the degree value of collision is less than the threshold The training sample of value is labeled as negative sample;Using logistic regression algorithm, according to the driving behavior data, logic is carried out to the positive sample and the negative sample and is returned Return training, obtain the driving behavior analysis model.
- 22. the driving behavior analysis device according to claim 20 or 21, which is characterized in that further include test cell, fit In selecting the training sample of preset number at random from the training sample as test sample;It is verified using the test sample Accuracy validation is carried out to the driving behavior analysis model.
- 23. driving behavior analysis device according to claim 22, which is characterized in that the test cell, suitable for by institute It states test sample and is randomly divided into N groups test subsample, N is natural number, and N >=3;Using wherein arbitrary N-1 groups test subsample Accuracy validation is carried out to one group of test subsample except the N-1 groups, until N groups test subsample is verified.
- 24. driving behavior analysis device according to claim 15, which is characterized in that the computing unit, suitable for obtaining Adjusting thresholds coefficient;According to the adjusting thresholds coefficient, each driving is obtained from the driving behavior analysis model trained Weight corresponding to behavior respectively, wherein, in the driving behavior analysis model, under different adjusting thresholds coefficients, respectively drive It sails behavior and corresponds to different weights in first mapping relations respectively;Each driving behavior according to getting exists respectively Corresponding weight in first mapping relations calculates the collision probability of happening of the vehicle.
- 25. driving behavior analysis device according to claim 15, which is characterized in that further include:Updating unit, suitable for When monitoring preset model modification trigger event, the driving behavior data for obtaining the vehicle are used as more new samples, using institute It states more new samples and training is updated to the driving behavior analysis model, and the driving behavior analysis mould trained and obtained will be updated Type is as the driving behavior analysis model trained.
- 26. a kind of driving behavior analysis system, which is characterized in that including:In data acquisition device and such as claim 15 to 25 Any one of them driving behavior analysis device, wherein:The data acquisition device, suitable for obtain driving behavior data of the vehicle in default first duration and it is described first when The driving range of the vehicle in length.
- 27. driving behavior analysis system according to claim 26, which is characterized in that the data acquisition device is suitable for Driving behavior data of the vehicle in default first duration are acquired by the driving states sensor installed on the vehicle, Or driving behavior data of the vehicle in default first duration are obtained from vehicle storage device.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611262696.0A CN108268678B (en) | 2016-12-30 | 2016-12-30 | Driving behavior analysis method, device and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611262696.0A CN108268678B (en) | 2016-12-30 | 2016-12-30 | Driving behavior analysis method, device and system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108268678A true CN108268678A (en) | 2018-07-10 |
CN108268678B CN108268678B (en) | 2021-07-23 |
Family
ID=62755268
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201611262696.0A Active CN108268678B (en) | 2016-12-30 | 2016-12-30 | Driving behavior analysis method, device and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108268678B (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110143202A (en) * | 2019-04-09 | 2019-08-20 | 南京交通职业技术学院 | A kind of dangerous driving identification and method for early warning and system |
CN113968234A (en) * | 2021-11-29 | 2022-01-25 | 蔡定海 | Vehicle auxiliary driving control method and device and vehicle-mounted controller |
Citations (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20130135092A1 (en) * | 2011-11-25 | 2013-05-30 | National Chiao Tung University | Driving behavior analysis and warning system and method thereof |
CN103310202A (en) * | 2013-06-27 | 2013-09-18 | 西安电子科技大学 | System and method for guaranteeing driving safety |
CN104050551A (en) * | 2014-05-23 | 2014-09-17 | 北京中交兴路信息科技有限公司 | Intelligent cargo allocation method and system based on prediction of destinations of vehicles |
CN104063914A (en) * | 2014-07-01 | 2014-09-24 | 深圳市成为智能交通系统有限公司 | Terminal achieving driving risk management and use method thereof |
CN104200267A (en) * | 2014-09-23 | 2014-12-10 | 清华大学 | Vehicle driving economy evaluation system and vehicle driving economy evaluation method |
CN104494601A (en) * | 2014-12-18 | 2015-04-08 | 清华大学苏州汽车研究院(吴江) | Driving behavior analysis and driving assistance system based on OBD |
US20150120336A1 (en) * | 2013-10-24 | 2015-04-30 | Tourmaline Labs, Inc. | Systems and methods for collecting and transmitting telematics data from a mobile device |
CN104978777A (en) * | 2014-04-11 | 2015-10-14 | 比亚迪股份有限公司 | System, device and method for analyzing driving behavior |
CN105046765A (en) * | 2015-08-19 | 2015-11-11 | 莆田市云驰新能源汽车研究院有限公司 | Automobile data recorder and method for improving driving behavior based on automobile data recorder |
CN105488046A (en) * | 2014-09-16 | 2016-04-13 | 钛马信息网络技术有限公司 | Big data analysis system based on vehicle insurance services |
CN105679020A (en) * | 2016-01-29 | 2016-06-15 | 深圳市美好幸福生活安全系统有限公司 | Driving behavior analysis device and method |
CN106095963A (en) * | 2016-06-17 | 2016-11-09 | 上海经达信息科技股份有限公司 | Vehicle drive behavior analysis big data public service platform under the Internet+epoch |
CN106240571A (en) * | 2016-08-03 | 2016-12-21 | 东软集团股份有限公司 | Driving behavior analysis method and apparatus |
-
2016
- 2016-12-30 CN CN201611262696.0A patent/CN108268678B/en active Active
Patent Citations (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20130135092A1 (en) * | 2011-11-25 | 2013-05-30 | National Chiao Tung University | Driving behavior analysis and warning system and method thereof |
CN103310202A (en) * | 2013-06-27 | 2013-09-18 | 西安电子科技大学 | System and method for guaranteeing driving safety |
US20150120336A1 (en) * | 2013-10-24 | 2015-04-30 | Tourmaline Labs, Inc. | Systems and methods for collecting and transmitting telematics data from a mobile device |
CN104978777A (en) * | 2014-04-11 | 2015-10-14 | 比亚迪股份有限公司 | System, device and method for analyzing driving behavior |
CN104050551A (en) * | 2014-05-23 | 2014-09-17 | 北京中交兴路信息科技有限公司 | Intelligent cargo allocation method and system based on prediction of destinations of vehicles |
CN104063914A (en) * | 2014-07-01 | 2014-09-24 | 深圳市成为智能交通系统有限公司 | Terminal achieving driving risk management and use method thereof |
CN105488046A (en) * | 2014-09-16 | 2016-04-13 | 钛马信息网络技术有限公司 | Big data analysis system based on vehicle insurance services |
CN104200267A (en) * | 2014-09-23 | 2014-12-10 | 清华大学 | Vehicle driving economy evaluation system and vehicle driving economy evaluation method |
CN104494601A (en) * | 2014-12-18 | 2015-04-08 | 清华大学苏州汽车研究院(吴江) | Driving behavior analysis and driving assistance system based on OBD |
CN105046765A (en) * | 2015-08-19 | 2015-11-11 | 莆田市云驰新能源汽车研究院有限公司 | Automobile data recorder and method for improving driving behavior based on automobile data recorder |
CN105679020A (en) * | 2016-01-29 | 2016-06-15 | 深圳市美好幸福生活安全系统有限公司 | Driving behavior analysis device and method |
CN106095963A (en) * | 2016-06-17 | 2016-11-09 | 上海经达信息科技股份有限公司 | Vehicle drive behavior analysis big data public service platform under the Internet+epoch |
CN106240571A (en) * | 2016-08-03 | 2016-12-21 | 东软集团股份有限公司 | Driving behavior analysis method and apparatus |
Non-Patent Citations (3)
Title |
---|
JU-TAEK OH: "Driving behavior analysis of commercial vehicles (buses) using a risky driving judgment device", 《2012 INTERNATIONAL JOURNAL OF HIGHWAY ENGINEERING 》 * |
胡水燕: "基于事故数据的道路车辆碰撞协调性研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 * |
雷虎: "愤怒情绪下的汽车驾驶行为特征及其对交通安全的影响研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110143202A (en) * | 2019-04-09 | 2019-08-20 | 南京交通职业技术学院 | A kind of dangerous driving identification and method for early warning and system |
CN113968234A (en) * | 2021-11-29 | 2022-01-25 | 蔡定海 | Vehicle auxiliary driving control method and device and vehicle-mounted controller |
CN113968234B (en) * | 2021-11-29 | 2023-05-02 | 深圳市科莱德电子有限公司 | Vehicle auxiliary driving control method and device and vehicle-mounted controller |
Also Published As
Publication number | Publication date |
---|---|
CN108268678B (en) | 2021-07-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
EP2710572B1 (en) | Vehicle data analysis method and vehicle data analysis system | |
CN109670970B (en) | Driving behavior scoring method and device and computer readable storage medium | |
CN110615001B (en) | Driving safety reminding method, device and medium based on CAN data | |
CN108877213B (en) | Traffic accident identification method and system | |
CN106097709B (en) | Driving behavior recognition methods based on intelligent vehicle mounted terminal | |
CN106650157B (en) | Method, device and system for estimating fault occurrence probability of vehicle parts | |
Castignani et al. | An evaluation study of driver profiling fuzzy algorithms using smartphones | |
CN111428960B (en) | Intelligent vehicle driving automatic evaluation method fusing multi-source vehicle-mounted sensor information | |
CN107953888B (en) | Road type identification method and system | |
US20130184928A1 (en) | Driver behavior diagnostic method and system | |
CN106781454B (en) | The appraisal procedure and device of driving behavior | |
DE102014006319A1 (en) | System for assessing and / or optimizing the operating behavior of a vehicle | |
US8983710B2 (en) | Methods and systems for determining driving performance variables | |
CN106022846A (en) | Automobile insurance pricing method, second-hand automobile pricing method and corresponding devices | |
CN112937591B (en) | Driving safety monitoring method, device, equipment and computer readable storage medium | |
CN106956680A (en) | A kind of electric automobile driving behavior identifying and analyzing method | |
CN113247008A (en) | Driving behavior monitoring method and device and electronic equipment | |
CN107918826A (en) | The driver's evaluation and dispatching method that a kind of driving environment perceives | |
CN110033617A (en) | A kind of train tracing model assessment system and method towards natural driving data | |
Cho et al. | DTG big data analysis for fuel consumption estimation | |
CN108268678A (en) | Driving behavior analysis method, apparatus and system | |
CN106651133B (en) | Method for achieving user driving behavior scoring based on ADAS by using mobile phone | |
CN115841712A (en) | Driving data processing method, device and equipment based on V2X technology | |
Higgs | Application of naturalistic truck driving data to analyze and improve car following models | |
CN115743143A (en) | Vehicle running condition analysis method and system based on Internet of vehicles |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |