CN115510990A - Model training method and related device - Google Patents
Model training method and related device Download PDFInfo
- Publication number
- CN115510990A CN115510990A CN202211234579.9A CN202211234579A CN115510990A CN 115510990 A CN115510990 A CN 115510990A CN 202211234579 A CN202211234579 A CN 202211234579A CN 115510990 A CN115510990 A CN 115510990A
- Authority
- CN
- China
- Prior art keywords
- vehicle
- vehicles
- driving
- determining
- driving characteristics
- 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.)
- Pending
Links
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
- B60W50/00—Details of control systems for road vehicle drive control not related to the control of a particular sub-unit, e.g. process diagnostic or vehicle driver interfaces
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
- B60W50/00—Details of control systems for road vehicle drive control not related to the control of a particular sub-unit, e.g. process diagnostic or vehicle driver interfaces
- B60W50/0097—Predicting future conditions
Landscapes
- Engineering & Computer Science (AREA)
- Automation & Control Theory (AREA)
- Human Computer Interaction (AREA)
- Transportation (AREA)
- Mechanical Engineering (AREA)
- Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)
Abstract
The application provides a model training method and a related device.A processing device can firstly acquire historical driving data and vehicle danger information which respectively correspond to a plurality of vehicles in a preset historical time period, wherein the vehicle danger information is used for identifying whether the plurality of vehicles are in danger or not in the preset historical time period. The processing device may determine a plurality of driving characteristics corresponding to the plurality of vehicles, respectively, based on the historical driving data. The processing device can determine the relevant parameters corresponding to the target driving characteristics, and select effective driving characteristics with reasonable relevance to risk analysis from a plurality of driving characteristics based on the relevant parameters, so that an emergence prediction model can be obtained through training based on the characteristics, the emergence probability of the vehicle can be accurately predicted through the emergence prediction model, the risk probability of the driver can be further reduced according to the dimension of analysis of the driving behavior of the driver, and the driving safety is improved.
Description
Technical Field
The present application relates to the field of model training technologies, and in particular, to a model training method and a related apparatus.
Background
The trip through the vehicle is the trip selection of most travelers at present, in order to ensure user's driving safety, need can carry out accurate analysis to the risk of traveling of vehicle.
In the related art, a method capable of accurately analyzing the vehicle driving risk is lacked, so that safety guarantee cannot be provided for the operation of the vehicle.
Disclosure of Invention
In order to solve the technical problem, the application provides a model training method, and the risk prediction model obtained by training in the mode can effectively predict the probability of vehicle risk.
The embodiment of the application discloses the following technical scheme:
in a first aspect, an embodiment of the present application discloses a model training method, where the method includes:
acquiring historical driving data and vehicle danger information which correspond to a plurality of vehicles in a preset historical time period respectively, wherein the vehicle danger information is used for identifying whether the plurality of vehicles are in danger in the preset historical time period;
determining a plurality of driving characteristics corresponding to the plurality of vehicles respectively based on the historical driving data, wherein the driving characteristics are used for identifying driving behavior characteristics of drivers corresponding to the plurality of vehicles respectively;
respectively taking the plurality of driving characteristics as target driving characteristics, and determining associated parameters corresponding to the target driving characteristics, wherein the associated parameters are used for identifying the degree of association between the target driving characteristics and the vehicle in danger or not;
determining effective driving characteristics of which associated parameters meet a preset parameter range in the plurality of driving characteristics;
and taking the effective driving characteristics corresponding to the vehicles as training samples, taking the vehicle risk information corresponding to the vehicles as training labels, and training to obtain a risk prediction model, wherein the risk prediction model is used for predicting the risk probability of the vehicles.
In a possible implementation manner, the determining of the associated parameter corresponding to the target driving feature includes;
according to the target driving characteristics, grouping the vehicles to obtain a plurality of vehicle groups;
determining evidence weights respectively corresponding to the plurality of vehicle groups;
and determining the information value corresponding to the target driving characteristics based on the evidence weights respectively corresponding to the plurality of vehicle groups, and determining the information value as the associated parameters corresponding to the target driving characteristics.
In one possible implementation manner, the determining the evidence weights respectively corresponding to the plurality of vehicle groups includes:
respectively taking the plurality of vehicle groups as target vehicle groups, and determining a first number of vehicles which are not in danger and are identified by corresponding vehicle danger information in the target vehicle groups, and a second number of vehicles which are in danger and are identified by the corresponding vehicle danger information;
determining a third number of vehicles which are not in danger and identified by corresponding vehicle danger information in the plurality of vehicles, and determining a fourth number of vehicles which are in danger and identified by corresponding vehicle danger information;
and determining the evidence weight corresponding to the target vehicle group according to a first ratio of the first number to the third number and a second ratio of the second number to the fourth number.
In one possible implementation, the evidence weight is calculated by the following formula:
where WOE is the evidence weight, y i Is a first number, y t Is a second number, n i Is a third number, n t Is the fourth number.
In one possible implementation manner, the determining the information value corresponding to the target driving feature based on the evidence weights respectively corresponding to the plurality of vehicle groups includes:
determining a sub-information value corresponding to the target vehicle group according to the evidence weight corresponding to the target vehicle group, the first ratio and the second ratio;
and determining the sum of the sub information values respectively corresponding to the plurality of vehicle groups as the information value corresponding to the target driving feature.
In one possible implementation, the method further includes:
calculating feature similarity between the plurality of driving features;
removing any of the two driving features from the plurality of driving features in response to a feature similarity between any two of the plurality of driving features being greater than a preset threshold.
In a second aspect, an embodiment of the present application discloses a model training apparatus, where the apparatus includes a first obtaining unit, a first determining unit, a second determining unit, a third determining unit, and a training unit:
the first acquisition unit is used for acquiring historical driving data and vehicle insurance information which correspond to a plurality of vehicles in a preset historical time period respectively, and the vehicle insurance information is used for identifying whether the plurality of vehicles are dangerous in the preset historical time period;
the first determination unit is configured to determine, based on the historical driving data, a plurality of driving characteristics corresponding to the plurality of vehicles, respectively, the driving characteristics being used to identify driving behavior characteristics of drivers corresponding to the plurality of vehicles, respectively;
the second determining unit is used for respectively taking the plurality of driving characteristics as target driving characteristics and determining associated parameters corresponding to the target driving characteristics, wherein the associated parameters are used for identifying the degree of association between the target driving characteristics and whether the vehicle is in danger or not;
the third determining unit is used for determining effective driving characteristics of which the associated parameters meet a preset parameter range in the plurality of driving characteristics;
the training unit is used for training to obtain an emergence prediction model by taking the effective driving characteristics corresponding to the vehicles as training samples and taking the vehicle emergence information corresponding to the vehicles as training labels, and the emergence prediction model is used for predicting the emergence probability of the vehicles.
In a possible implementation manner, the second determining unit is specifically configured to determine the second threshold value;
according to the target driving characteristics, grouping the vehicles to obtain a plurality of vehicle groups;
determining evidence weights respectively corresponding to the plurality of vehicle groups;
and determining the information value corresponding to the target driving characteristics based on the evidence weights respectively corresponding to the plurality of vehicle groups, and determining the information value as the associated parameters corresponding to the target driving characteristics.
In a possible implementation manner, the second determining unit is specifically configured to:
respectively taking the plurality of vehicle groups as target vehicle groups, and determining a first number of vehicles which are not in danger and are identified by corresponding vehicle danger information in the target vehicle groups, and a second number of vehicles which are in danger and are identified by the corresponding vehicle danger information;
determining a third number of vehicles in the plurality of vehicles which are not in danger and identified by corresponding vehicle danger information, and determining a fourth number of vehicles in danger and identified by corresponding vehicle danger information;
and determining the evidence weight corresponding to the target vehicle group according to a first ratio of the first number to the third number and a second ratio of the second number to the fourth number.
In one possible implementation, the evidence weight is calculated by the following formula:
where WOE is the evidence weight, y i Is a first numberAmount, y t Is a second number, n i Is a third number, n t Is the fourth number.
In a possible implementation manner, the second determining unit is specifically configured to:
determining a sub-information value corresponding to the target vehicle group according to the evidence weight corresponding to the target vehicle group, the first ratio and the second ratio;
and determining the sum of the sub information values respectively corresponding to the plurality of vehicle groups as the information value corresponding to the target driving characteristic.
In one possible implementation, the apparatus further includes a computing unit and a response unit:
the calculation unit is used for calculating feature similarity among the plurality of driving features;
the response unit is used for removing any one of the two driving characteristics from the plurality of driving characteristics in response to the characteristic similarity between any two of the plurality of driving characteristics being larger than a preset threshold value.
In a third aspect, an embodiment of the present application discloses a computer device, where the computer device includes a processor and a memory:
the memory is used for storing program codes and transmitting the program codes to the processor;
the processor is configured to perform the model training method of any one of the first aspect according to instructions in the program code.
In a fourth aspect, an embodiment of the present application discloses a computer-readable storage medium for storing a computer program, where the computer program is used to execute the model training method in any one of the first aspect.
In a fifth aspect, embodiments of the present application disclose a computer program product comprising instructions, which when run on a computer, cause the computer to perform the model training method of any one of the first aspect.
According to the technical scheme, in order to train and obtain the model capable of accurately predicting the vehicle driving risk, the processing device may first obtain historical driving data and vehicle insurance information respectively corresponding to the plurality of vehicles in a preset historical time period, wherein the vehicle insurance information is used for identifying whether the plurality of vehicles are in insurance in the preset historical time period. The processing device may determine, based on the historical driving data, a plurality of driving characteristics corresponding to the plurality of vehicles, respectively, the driving characteristics identifying driving behavior characteristics of drivers corresponding to the plurality of vehicles, respectively. The processing device may determine a correlation parameter corresponding to the target driving characteristic, where the correlation parameter is used to identify a degree of correlation between the target driving characteristic and whether the vehicle is in danger or not. The processing equipment can select effective driving characteristics with reasonable relevance to risk analysis from the multiple driving characteristics based on the relevant parameters, so that an emergency prediction model can be obtained through training based on the characteristics, the emergency probability of the vehicle can be accurately predicted through the emergency prediction model, the risk probability of the driver can be further reduced through the dimension of analysis of the driving behavior of the driver, and the driving safety is improved.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
Fig. 1 is a flowchart of a model training method provided in an embodiment of the present application;
fig. 2 is a schematic diagram of a model training method in an actual application scenario according to an embodiment of the present application;
fig. 3 is a schematic diagram of a model training method in an actual application scenario according to an embodiment of the present application;
fig. 4 is a block diagram of a model training apparatus according to an embodiment of the present disclosure.
Detailed Description
Embodiments of the present application are described below with reference to the accompanying drawings.
It is understood that the method may be applied to a processing device, which is a processing device capable of performing model training, for example, a terminal device or a server with a model training function. The method can be independently executed through the terminal equipment or the server, can also be applied to a network scene of communication between the terminal equipment and the server, and can be executed through the cooperation of the terminal equipment and the server. The terminal device may be a computer, a mobile phone, or the like. The server may be understood as an application server or a Web server, and in actual deployment, the server may be an independent server or a cluster server.
Referring to fig. 1, fig. 1 is a flowchart of a model training method provided in an embodiment of the present application, where the method includes:
s101: historical driving data and vehicle insurance information which correspond to a plurality of vehicles in a preset historical time period are obtained.
The vehicle insurance information is used for identifying whether a plurality of vehicles are in insurance or not within a preset historical period, the historical driving data is used for identifying the driving conditions of the plurality of vehicles in the preset historical period, and the preset historical period can be flexibly set based on actual demands, for example, the preset historical period can be 6 months in the past.
S102: based on the historical driving data, a plurality of driving characteristics corresponding to the plurality of vehicles, respectively, are determined.
The processing device may determine, from the historical driving data, a plurality of driving characteristics corresponding to each vehicle in the preset historical period, the driving characteristics identifying driving behavior characteristics of drivers corresponding to the plurality of vehicles, respectively, and the driving characteristics may include, for example, a driving mileage characteristic, an over-speed driving characteristic, a driving duration characteristic, a peak driving characteristic in the morning and evening, a driving night characteristic, a fatigue driving characteristic, and the like. The driving mileage is characterized by an average mileage of each driving of the vehicle in a preset history period, the driving period is characterized by an average period of each driving of the vehicle in the preset history period, the overspeed driving is characterized by a proportion of all driving of the vehicle in which an overspeed condition occurs in the preset history period, the rush hour driving is characterized by a proportion of all driving of the vehicle in which the vehicle peaks in the morning and evening in the preset history period, the night driving is characterized by a proportion of all driving of the vehicle in which the vehicle travels at night in the preset history period, and the fatigue driving is characterized by a proportion of all driving of the vehicle in which a fatigue driving condition occurs in the preset history period.
S103: and respectively taking the plurality of driving characteristics as target driving characteristics, and determining the associated parameters corresponding to the target driving characteristics.
In order to determine the characteristics that can be used for accurately analyzing the vehicle risk occurrence probability, the processing device may first analyze the association between each characteristic and whether the vehicle is in risk, and the association parameter is used for identifying the degree of association between the target driving characteristic and whether the vehicle is in risk.
S104: and determining effective driving characteristics of which the associated parameters meet a preset parameter range in the plurality of driving characteristics.
The processing device may preset a parameter range for selecting driving characteristics with reasonable relevance. For example, the processing device may remove driving characteristics with very high relevance, which are not true enough; the driving features with very low relevance to the risk probability can be removed, and the prediction of the occurrence probability is less helpful for the part of the driving features.
S105: and taking the effective driving characteristics corresponding to the vehicles as training samples, taking the vehicle risk information corresponding to the vehicles as training labels, and training to obtain a risk prediction model.
Wherein the risk prediction model is used for predicting the risk probability of the vehicle. The specific training process may be as follows:
the processing device can obtain an initial risk prediction model, and input effective driving characteristics corresponding to a certain vehicle into the model to obtain undetermined risk information, wherein the undetermined risk information is whether the vehicle predicted by the initial risk prediction model is in risk or not. The processing device can adjust the model parameters of the initial risk prediction model according to the difference between the undetermined risk information and the vehicle risk information corresponding to the vehicle, so that the undetermined risk information predicted by the processing device is matched with the vehicle risk information. Therefore, the model trained by the method can effectively predict whether the vehicle will take the risk or not based on the effective driving characteristics, and the risk taking probability is obtained.
According to the technical scheme, in order to train and obtain the model capable of accurately predicting the vehicle driving risk, the processing device can firstly obtain historical driving data and vehicle danger information which correspond to the plurality of vehicles in the preset historical time period respectively, and the vehicle danger information is used for identifying whether the plurality of vehicles are in danger or not in the preset historical time period. The processing device may determine, based on the historical driving data, a plurality of driving characteristics corresponding to the plurality of vehicles, respectively, the driving characteristics identifying driving behavior characteristics of drivers corresponding to the plurality of vehicles, respectively. The processing device may determine a correlation parameter corresponding to the target driving characteristic, where the correlation parameter is used to identify a degree of correlation between the target driving characteristic and whether the vehicle is in danger or not. The processing equipment can select effective driving characteristics with reasonable relevance to risk analysis from the multiple driving characteristics based on the relevant parameters, so that an emergency prediction model can be obtained through training based on the characteristics, the emergency probability of the vehicle can be accurately predicted through the emergency prediction model, the risk probability of the driver can be further reduced through the dimension of analysis of the driving behavior of the driver, and the driving safety is improved.
In one possible implementation, the processing device may calculate the associated parameter corresponding to the target driving characteristic in the following manner.
The processing device may perform grouping processing on the plurality of vehicles according to the target driving characteristics, resulting in a plurality of vehicle groups. For example, when the target driving characteristic is a driving range characteristic, the vehicle grouping may be performed based on mileage.
The processing device may determine evidence weights respectively corresponding to the plurality of vehicle groups, and the evidence weights are used for reflecting the relevance degree of the target driving feature to whether the vehicle is in danger in a certain group. The processing device may determine an information value corresponding to the target driving feature based on the evidence weights respectively corresponding to the plurality of vehicle groups, and determine the information value as the associated parameter corresponding to the target driving feature. Therefore, the processing equipment can analyze the relevance of each driving characteristic to the risk probability in a fine-grained manner, and obtain more accurate effective driving characteristics.
In one possible implementation, when determining the evidence weights respectively corresponding to the plurality of vehicle groups, the processing device may respectively regard the plurality of vehicle groups as target vehicle groups, determine a first number of vehicles in the target vehicle groups for which the corresponding vehicle insurance information is identified as not in insurance, and determine a second number of vehicles in insurance.
The processing device may then determine a third number of vehicles of the plurality of vehicles for which corresponding vehicle insurance information identifies an unfulfilled vehicle and a fourth number of vehicles for which corresponding vehicle insurance information identifies an insured vehicle.
The proportion of the vehicles in the target vehicle group that do not go out of danger in the entire vehicles in the target vehicle group that do not go out of danger can be reflected by a first ratio of the first number to the third number, and the proportion of the vehicles in the target vehicle group that go out of danger in the entire vehicles in the target vehicle group that go out of danger can be reflected by a second ratio of the second number to the fourth number, so the degree of association between the corresponding driving characteristics in the target vehicle group and whether or not to go out of danger can be reflected by the two ratios.
Based on this, the processing device may determine the evidence weight corresponding to the target vehicle group according to a first ratio of the first number to the third number, and a second ratio of the second number to the fourth number.
In one possible implementation, the evidence weight is calculated by the following formula:
where WOE is the evidence weight, y i Is a first number, y t Is a second numberAmount, n i Is a third number, n t Is the fourth number.
In a possible implementation manner, when determining the information value corresponding to the target driving feature based on the evidence weights respectively corresponding to the plurality of vehicle groups, the processing device may determine, according to the evidence weight corresponding to the target vehicle group, the first ratio and the second ratio, a sub-information value corresponding to the target vehicle group, where the sub-information value is used to identify a degree of association of the target driving feature with respect to an occurrence probability in the target vehicle group.
The processing device may determine a sum of sub information values respectively corresponding to the plurality of vehicle groups as the target driving characteristic corresponding information value.
As shown in the following equation:
wherein n is the number of vehicle groups IV i And IV is the information value corresponding to the driving characteristics.
It can be understood that when the similarity between the two characteristics is high, for example, the driving duration characteristic and the driving mileage characteristic are reflected to a certain extent by the mileage of the driver in a single driving, the association degree of the driving duration characteristic and the driving mileage characteristic to the risk probability is similar, and the effect on the prediction of the risk probability to the vehicle is also similar. Therefore, in order to improve the processing efficiency of the model, in a possible implementation manner, the model prediction may be performed on only one of the features of the two feature processing devices with higher similarity.
The processing device may calculate a feature similarity between the plurality of driving features and determine whether the feature similarity between two features is greater than a preset threshold. In response to a feature similarity between any two of the plurality of driving features being greater than a preset threshold, the processing device removes any one of the two driving features from the plurality of driving features. For example, the processing device may retain the feature with the greater associated parameter of the two features.
In order to facilitate understanding of the technical solution provided by the embodiment of the present application, a model training method provided by the embodiment of the present application will be introduced in combination with an actual application scenario.
Referring to fig. 2, fig. 2 is a schematic view of a model training method in an actual application scenario provided by the embodiment of the present application, a processing device may acquire driving trajectory data through a vehicle-to-vehicle machine of a vehicle network, and acquire an emergence record of a corresponding time period of the vehicle, where the driving trajectory data is historical driving data, and the emergence record is vehicle emergence information.
Data preprocessing: the original data are cleaned, processed by abnormal values and divided into journey data, and then the journey data are processed into driving behavior characteristics, including driving mileage characteristics, overspeed driving characteristics, driving duration characteristics, peak driving characteristics in the morning and evening, night driving characteristics, fatigue driving characteristics and the like.
Carrying out feature discretization treatment: feature binning and WOE conversion.
WOE is calculated as
y i Is the number of no risk in the set of samples, n i Is the number of non-occurrences in the total sample, y t Is the number of occurrences in the set of samples, n t Is the number of risks in the total sample.
Selecting characteristics: and calculating the correlation and IV value of the features, filtering out the features with higher correlation, retaining the feature with higher IV value when the correlation exceeds 0.8, and removing the feature with IV < 0.02.
The IV value is calculated by the formula
n is the number of feature packets.
Characteristic normalization: and carrying out normalization processing on the selected features.
The processing device can use the processed driving characteristics as samples, and the corresponding risk records as sample labels to train a logistic regression model to obtain model parameters.
After the model is obtained through training, the processing equipment can process the travel data of the vehicle to be predicted into target features, the target features are subjected to box separation, WOE conversion and normalization processing through the same method in the training process, the risk probability y is predicted through the trained logistic regression model, and the driving risk is divided into 5 grades according to the predicted value. With the place value as the endpoint of the hierarchy, the division is as shown in the following table
| Score of | Class of risk classification | Degree of risk |
| y is less than or equal to 0.1 quantile value | 1 | Is very low |
| Value of 0.1 quantile<y is less than or equal to 0.35 quantile value | 2 | Lower is |
| Value of 0.35 quantile<y is less than or equal to 0.65 |
3 | In general terms |
| Value of 0.65 quantile<y is less than or equal to 0.9 quantile value | 4 | Is higher than |
| y>Value of 0.9 |
5 | Is very high |
: 2398 travel data of 6 months of vehicles and insurance records corresponding to the travel data are obtained.
Data preprocessing: 16 features such as driving range/ratio over speed/early peak driving range/night driving range/fatigue driving range/daily average driving range are generated based on trip data of sample vehicles (1918).
Carrying out feature discretization treatment: the features are segmented with optimal binning and then WOE transformed.
Selecting characteristics: calculating the characteristic correlation and IV value, wherein the characteristic with the correlation exceeding 0.8 retains the larger IV value, and the characteristic with the IV less than 0.02 is removed.
Characteristic normalization: the features are normalized.
Training a logistic regression model: and (4) taking the processed driving characteristics as a sample, and taking the corresponding risk record as a sample label to train a logistic regression model.
The practical application is as follows: processing the travel data of the vehicles to be tested (480) into target characteristics, performing segmentation and WOE conversion on the target characteristics by adopting the same method, then performing standardization processing by adopting the same method, then predicting the risk probability of the vehicles by using the trained model, and finally dividing the risk grades according to the above table to obtain the risk rates of all grades as shown in figure 3. The rate of emergence for five grades of vehicles was 4.17%, 5.00%, 11.81%, 20.00%, 31.25%, respectively. The risk degree of the third-level driving is general, and the risk rate of the third-level driving is close to the average level of 13.33%; the corresponding risk ratio of low driving risk of the first two levels is also lower; the fifth level has the highest driving risk, and the risk rate is far higher than the average level.
Based on the model training method provided in the foregoing embodiment, an embodiment of the present application further provides a model training device, referring to fig. 4, and fig. 4 is a block diagram of a structure of the model training device provided in the embodiment of the present application, where the device includes a first obtaining unit 401, a first determining unit 402, a second determining unit 403, a third determining unit 404, and a training unit 405:
the first obtaining unit 401 is configured to obtain historical driving data and vehicle insurance information respectively corresponding to a plurality of vehicles in a preset historical time period, where the vehicle insurance information is used to identify whether the plurality of vehicles are in insurance in the preset historical time period;
the first determining unit 402 is configured to determine, based on the historical driving data, a plurality of driving characteristics corresponding to the plurality of vehicles, respectively, where the driving characteristics are used to identify driving behavior characteristics of drivers corresponding to the plurality of vehicles, respectively;
the second determining unit 403 is configured to determine, by taking the multiple driving characteristics as target driving characteristics respectively, associated parameters corresponding to the target driving characteristics, where the associated parameters are used to identify a degree of association between the target driving characteristics and whether the vehicle is in danger or not;
the third determining unit 404 is configured to determine an effective driving characteristic of the plurality of driving characteristics, where the associated parameter meets a preset parameter range;
the training unit 405 is configured to train to obtain an emergence prediction model by using the effective driving features corresponding to the multiple vehicles as training samples and using the vehicle emergence information corresponding to the multiple vehicles as training labels, where the emergence prediction model is used to predict the emergence probability of the vehicle.
In a possible implementation manner, the second determining unit 403 is specifically configured to;
according to the target driving characteristics, grouping the vehicles to obtain a plurality of vehicle groups;
determining evidence weights respectively corresponding to the plurality of vehicle groups;
and determining the information value corresponding to the target driving characteristics based on the evidence weights respectively corresponding to the plurality of vehicle groups, and determining the information value as the associated parameters corresponding to the target driving characteristics.
In a possible implementation manner, the second determining unit 403 is specifically configured to:
respectively taking the plurality of vehicle groups as target vehicle groups, and determining a first number of vehicles which are not in danger and are identified by corresponding vehicle danger information in the target vehicle groups, and a second number of vehicles which are in danger and are identified by the corresponding vehicle danger information;
determining a third number of vehicles in the plurality of vehicles which are not in danger and identified by corresponding vehicle danger information, and determining a fourth number of vehicles in danger and identified by corresponding vehicle danger information;
and determining the evidence weight corresponding to the target vehicle group according to a first ratio of the first number to the third number and a second ratio of the second number to the fourth number.
In one possible implementation, the evidence weight is calculated by the following formula:
where WOE is the evidence weight, y i Is a first number, y t Is a second number, n i Is a third number, n t Is the fourth number.
In a possible implementation manner, the second determining unit 403 is specifically configured to:
determining a sub-information value corresponding to the target vehicle group according to the evidence weight corresponding to the target vehicle group, the first ratio and the second ratio;
and determining the sum of the sub information values respectively corresponding to the plurality of vehicle groups as the information value corresponding to the target driving feature.
In one possible implementation, the apparatus further includes a computing unit and a response unit:
the calculation unit is used for calculating feature similarity among the plurality of driving features;
the response unit is used for removing any one of the two driving characteristics from the plurality of driving characteristics in response to the characteristic similarity between any two of the plurality of driving characteristics being larger than a preset threshold value.
The embodiment of the application also provides a computer device, and the processor included in the terminal device further has the following functions:
acquiring historical driving data and vehicle danger information which correspond to a plurality of vehicles in a preset historical time period respectively, wherein the vehicle danger information is used for identifying whether the plurality of vehicles are in danger in the preset historical time period;
determining a plurality of driving characteristics corresponding to the plurality of vehicles respectively based on the historical driving data, wherein the driving characteristics are used for identifying driving behavior characteristics of drivers corresponding to the plurality of vehicles respectively;
respectively taking the plurality of driving characteristics as target driving characteristics, and determining associated parameters corresponding to the target driving characteristics, wherein the associated parameters are used for identifying the degree of association between the target driving characteristics and the vehicle in danger or not;
determining effective driving characteristics of which associated parameters meet a preset parameter range in the plurality of driving characteristics;
and taking the effective driving characteristics corresponding to the vehicles as training samples, taking the vehicle risk information corresponding to the vehicles as training labels, and training to obtain a risk prediction model, wherein the risk prediction model is used for predicting the risk probability of the vehicles.
The units described in the embodiments of the present disclosure may be implemented by software or hardware. Where the name of an element does not in some cases constitute a limitation on the element itself.
In addition, a storage medium is further provided in an embodiment of the present application, where the storage medium is used to store a computer program, and the computer program is used to execute the model training method provided in the foregoing embodiment.
Embodiments of the present application further provide a computer program product including instructions, which when run on a computer, cause the computer to execute the model training method provided by the above embodiments.
Those of ordinary skill in the art will understand that: all or part of the steps for realizing the method embodiments can be completed by hardware related to program instructions, the program can be stored in a computer readable storage medium, and the program executes the steps comprising the method embodiments when executed; and the aforementioned storage medium may be at least one of the following media: various media that can store program codes, such as a read-only memory (ROM), a RAM, a magnetic disk, or an optical disk.
It should be noted that, in the present specification, all the embodiments are described in a progressive manner, and the same and similar parts among the embodiments may be referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, the apparatus and system embodiments, because they are substantially similar to the method embodiments, are described in a relatively simple manner, and reference may be made to some of the descriptions of the method embodiments for related points. The above-described embodiments of the apparatus and system are merely illustrative, and the units described as separate parts may or may not be physically separate, and the parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement without inventive effort.
The above description is only one specific embodiment of the present application, but the scope of the present application is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present application should be covered within the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.
Claims (10)
1. A method of model training, the method comprising:
acquiring historical driving data and vehicle danger information which correspond to a plurality of vehicles in a preset historical time period respectively, wherein the vehicle danger information is used for identifying whether the plurality of vehicles are in danger in the preset historical time period;
determining a plurality of driving characteristics corresponding to the plurality of vehicles respectively based on the historical driving data, wherein the driving characteristics are used for identifying driving behavior characteristics of drivers corresponding to the plurality of vehicles respectively;
respectively taking the plurality of driving characteristics as target driving characteristics, and determining associated parameters corresponding to the target driving characteristics, wherein the associated parameters are used for identifying the degree of association between the target driving characteristics and the vehicle in danger or not;
determining effective driving characteristics of which associated parameters meet a preset parameter range in the plurality of driving characteristics;
and taking the effective driving characteristics corresponding to the vehicles as training samples, taking the vehicle risk information corresponding to the vehicles as training labels, and training to obtain a risk prediction model, wherein the risk prediction model is used for predicting the risk probability of the vehicles.
2. The method according to claim 1, wherein the determining of the associated parameter corresponding to the target driving feature comprises;
according to the target driving characteristics, grouping the vehicles to obtain a plurality of vehicle groups;
determining evidence weights respectively corresponding to the plurality of vehicle groups;
and determining the information value corresponding to the target driving characteristics based on the evidence weights respectively corresponding to the plurality of vehicle groups, and determining the information value as the associated parameters corresponding to the target driving characteristics.
3. The method of claim 2, wherein the determining the evidence weights respectively corresponding to the plurality of vehicle groupings comprises:
respectively taking the plurality of vehicle groups as target vehicle groups, and determining a first number of vehicles which are not in danger and are identified by corresponding vehicle danger information in the target vehicle groups, and a second number of vehicles which are in danger and are identified by the corresponding vehicle danger information;
determining a third number of vehicles which are not in danger and identified by corresponding vehicle danger information in the plurality of vehicles, and determining a fourth number of vehicles which are in danger and identified by corresponding vehicle danger information;
and determining the evidence weight corresponding to the target vehicle group according to a first ratio of the first quantity to the third quantity and a second ratio of the second quantity to the fourth quantity.
4. The method according to claim 3, wherein the evidence weight is calculated by the following formula:
where WOE is the evidence weight, y i Is a first number, y t Is a second number, n i Is a third number, n t Is the fourth number.
5. The method according to claim 3, wherein the determining the information value corresponding to the target driving feature based on the evidence weights respectively corresponding to the plurality of vehicle groups comprises:
determining a sub-information value corresponding to the target vehicle group according to the evidence weight corresponding to the target vehicle group, the first ratio and the second ratio;
and determining the sum of the sub information values respectively corresponding to the plurality of vehicle groups as the information value corresponding to the target driving feature.
6. The method of claim 1, further comprising:
calculating feature similarity between the plurality of driving features;
removing any one of the two driving features from the plurality of driving features in response to a feature similarity between any two of the plurality of driving features being greater than a preset threshold.
7. A model training device is characterized by comprising a first acquisition unit, a first determination unit, a second determination unit, a third determination unit and a training unit:
the first acquisition unit is used for acquiring historical driving data and vehicle insurance information which correspond to a plurality of vehicles in a preset historical time period respectively, and the vehicle insurance information is used for identifying whether the plurality of vehicles are dangerous in the preset historical time period;
the first determination unit is configured to determine, based on the historical driving data, a plurality of driving characteristics corresponding to the plurality of vehicles, respectively, the driving characteristics being used to identify driving behavior characteristics of drivers corresponding to the plurality of vehicles, respectively;
the second determining unit is used for respectively taking the plurality of driving characteristics as target driving characteristics and determining associated parameters corresponding to the target driving characteristics, wherein the associated parameters are used for identifying the degree of association between the target driving characteristics and whether the vehicle is in danger or not;
the third determining unit is used for determining effective driving characteristics of which the associated parameters meet a preset parameter range in the plurality of driving characteristics;
the training unit is used for training to obtain an emergence prediction model by taking the effective driving characteristics corresponding to the vehicles as training samples and taking the vehicle emergence information corresponding to the vehicles as training labels, and the emergence prediction model is used for predicting the emergence probability of the vehicles.
8. A computer device, the computer device comprising a processor and a memory:
the memory is used for storing program codes and transmitting the program codes to the processor;
the processor is configured to perform the model training method of any one of claims 1-6 according to instructions in the program code.
9. A computer-readable storage medium for storing a computer program for performing the model training method of any one of claims 1-6.
10. A computer program product comprising instructions which, when run on a computer, cause the computer to perform the model training method of any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211234579.9A CN115510990A (en) | 2022-10-10 | 2022-10-10 | Model training method and related device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211234579.9A CN115510990A (en) | 2022-10-10 | 2022-10-10 | Model training method and related device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115510990A true CN115510990A (en) | 2022-12-23 |
Family
ID=84507817
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211234579.9A Pending CN115510990A (en) | 2022-10-10 | 2022-10-10 | Model training method and related device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115510990A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116091254A (en) * | 2023-04-11 | 2023-05-09 | 天津所托瑞安汽车科技有限公司 | Commercial vehicle risk analysis method |
| CN116226787A (en) * | 2023-05-04 | 2023-06-06 | 中汽信息科技(天津)有限公司 | Commercial vehicle danger probability prediction method, equipment and medium |
-
2022
- 2022-10-10 CN CN202211234579.9A patent/CN115510990A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116091254A (en) * | 2023-04-11 | 2023-05-09 | 天津所托瑞安汽车科技有限公司 | Commercial vehicle risk analysis method |
| CN116091254B (en) * | 2023-04-11 | 2023-08-01 | 天津所托瑞安汽车科技有限公司 | Commercial vehicle risk analysis method |
| CN116226787A (en) * | 2023-05-04 | 2023-06-06 | 中汽信息科技(天津)有限公司 | Commercial vehicle danger probability prediction method, equipment and medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115510990A (en) | Model training method and related device | |
| CN110675267B (en) | Method and system for carrying out vehicle insurance early warning according to real-time road conditions | |
| CN111209400B (en) | Data analysis method and device | |
| CN110705774A (en) | Vehicle energy consumption analysis prediction method and system | |
| WO2013184262A2 (en) | System and method for computing | |
| CN109636482B (en) | Data processing method and system based on similarity model | |
| CN112560465B (en) | Batch abnormal event monitoring method and device, electronic equipment and storage medium | |
| CN113762755A (en) | Method and device for pushing driver analysis report, computer equipment and storage medium | |
| CN110533094B (en) | Evaluation method and system for driver | |
| CN112232606B (en) | Service data prediction method, device, equipment and storage medium | |
| CN113379318A (en) | Method and device for evaluating operation service quality of public transport system and computer equipment | |
| CN113032239A (en) | Risk prompting method and device, electronic equipment and storage medium | |
| CN115114124A (en) | Host risk assessment method and device | |
| CN109035480B (en) | Report data generation method | |
| CN112950030B (en) | Residual error assessment method and device for electric automobile, electronic equipment and storage medium | |
| US20230260342A1 (en) | Method and computer programmes for the management of vehicle fleets | |
| CN115221218A (en) | Quality evaluation method and device for vehicle data, computer equipment and storage medium | |
| CN110796379B (en) | Risk assessment method, device and equipment of business channel and storage medium | |
| CN112398706B (en) | Data evaluation standard determining method and device, storage medium and electronic equipment | |
| CN113805564A (en) | Equipment fault early warning and diagnosis system based on CEP | |
| CN112927091A (en) | Complaint early warning method and device for annuity insurance, computer equipment and medium | |
| CN113011596A (en) | Method, device and system for automatically updating model and electronic equipment | |
| Siaminamini et al. | Generating a risk profile for car insurance policyholders: A deep learning conceptual model | |
| CN116665342B (en) | New energy automobile driving behavior analysis method, system and equipment | |
| CN117493820B (en) | Data element processing method and device |
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 | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20230718 Address after: No. 203, Shanghai Songjiang Road, No. 201563, Pudong New Area Applicant after: SAIC Motor Corp.,Ltd. Applicant after: Shanghai automotive industry (Group) Co.,Ltd. Address before: Room 509, building 1, 563 Songtao Road, China (Shanghai) pilot Free Trade Zone, Pudong New Area, Shanghai, 201203 Applicant before: SAIC Motor Corp.,Ltd. |