CN107766876B

CN107766876B - Driving model training method, driver's recognition methods, device, equipment and medium

Info

Publication number: CN107766876B
Application number: CN201710846242.6A
Authority: CN
Inventors: 吴壮伟; 金鑫; 张川
Original assignee: Ping An Technology Shenzhen Co Ltd
Current assignee: Ping An Technology Shenzhen Co Ltd
Priority date: 2017-09-19
Filing date: 2017-09-19
Publication date: 2019-08-13
Anticipated expiration: 2037-09-19
Also published as: CN107766876A; WO2019056470A1

Abstract

The present invention discloses a kind of driving model training method, driver's recognition methods, device, equipment and medium.The driving model training method includes: the training behavioral data for obtaining user, and the trained behavioral data is associated with user identifier；Based on the trained behavioral data, trained driving data associated with the user identifier is obtained；Based on the user identifier, positive negative sample is obtained from the trained driving data；Error backward propagation method model is trained using the positive negative sample, obtains target driving model.The driving model training method solves the problems, such as that current driving model recognition effect is poor, and improves the accuracy that identification driver drives.

Description

Driving model training method, driver's recognition methods, device, equipment and medium

Technical field

The present invention relates to data monitoring field more particularly to a kind of built-up pattern training method, drive recognition methods, dress It sets, equipment and medium.

Background technique

With the development of information age, it is raw that artificial intelligence technology is more and more made to solve people as core technology Particular problem in work.Currently, being carried out for whether I drives to calculate practical mileage problem using sole user's model in the industry Two classification outputs are to determine whether that I drives.Face the user group to grow stronger day by day, more and more users' model file, Biggish pressure can be caused to server storage.Instantly decision-tree model is used in the industry, directly uses input variable, only by Input variable itself is identified, is the demand for not being able to satisfy model, causes the accuracy rate for driving identification low.

Summary of the invention

The embodiment of the present invention provides a kind of driving model training method, device, equipment and medium, to solve current to drive mould The poor problem of type recognition effect.

The embodiment of the present invention also provides a kind of driver's recognition methods, device, equipment and medium, to solve current identification hand The lower problem of the accuracy that machine user drives.

In a first aspect, the embodiment of the present invention provides a kind of driving model training method, comprising:

The training behavioral data of user is obtained, the trained behavioral data is associated with user identifier；

Based on the trained behavioral data, trained driving data associated with the user identifier is obtained；

Based on the user identifier, positive negative sample is obtained from the trained driving data；

Error backward propagation method model is trained using the positive negative sample, obtains target driving model.

Second aspect, the embodiment of the present invention provide a kind of driving model training device, comprising:

Training behavioral data obtains module, for obtaining the training behavioral data of user, the trained behavioral data and use Family mark is associated；

Training driving data obtains module, for being based on the trained behavioral data, obtains related to the user identifier The training driving data of connection；

Positive and negative sample acquisition module obtains positive negative sample from the trained driving data for being based on the user identifier；

Target driving model obtain module, for using the positive negative sample to error backward propagation method model into Row training, obtains target driving model.

The third aspect, the embodiment of the present invention provide a kind of driver's recognition methods, comprising:

The behavioral data to be identified of user is obtained, the behavioral data to be identified is associated with user identifier；

Database is inquired based on the user identifier, obtains target driving model corresponding with the user identifier；

Based on the behavioral data to be identified and the target driving model, identification probability value is obtained；

Judge whether the identification probability value is greater than predetermined probabilities value；If the identification probability value is greater than the predetermined probabilities Value, it is determined that for my driving.

Fourth aspect, the embodiment of the present invention provide a kind of driver's identification device, comprising:

Behavioral data to be identified obtains module, for obtaining the behavioral data to be identified of user, the behavior number to be identified According to associated with user identifier；

Target driving model obtains module, for inquiring database based on the user identifier, obtains and marks with the user Sensible corresponding target driving model；

Identification probability value obtains module, for being based on the behavioral data to be identified and the target driving model, obtains Identification probability value；

Recognition result judgment module, for judging whether the identification probability value is greater than predetermined probabilities value；If the identification Probability value is greater than the predetermined probabilities value, it is determined that for my driving.

5th aspect, the embodiment of the present invention provide a kind of terminal device, including memory, processor and are stored in described In memory and the computer program that can run on the processor, the processor are realized when executing the computer program The step of driving model training method；Alternatively, the processor realizes the driver when executing the computer program The step of recognition methods.

6th aspect, the embodiment of the present invention provide a kind of computer-readable medium, and the computer-readable medium storage has The step of computer program, the computer program realizes driving model training method when being executed by processor；Alternatively, institute State the step of realizing driver's recognition methods when processor executes the computer program.

In driving model training method, device, equipment and medium provided by the embodiment of the present invention, the instruction of user is first obtained Practice behavioral data, training behavioral data is associated with user identifier, marks to obtain respectively based on user identifier with target user Know trained behavioral data corresponding with non-targeted user identifier, target can be identified with the target driving model for guaranteeing that training obtains The driving behavior of user.It is then based on trained behavioral data, obtains trained driving data associated with user identifier, the training Driving data is to exclude other non-driving behavior numbers never with the corresponding trained behavioral data of driving style is extracted in behavior type According to interference, advantageously ensure that training acquisition target driving model recognition accuracy and improve target driving model training Efficiency saves training duration.Then it is based on user identifier, obtains positive negative sample from training driving data, positive negative sample can have Parameter needed for determining training objective driving model is imitated, guarantees the accuracy for the target driving model recognition result that training obtains.Most Error backward propagation method model is trained using positive negative sample afterwards, obtains target driving model, it can effectively more The weight of each layer in new error backward propagation method model, so that being identified by the driving model that positive and negative sample training obtains Effect is more accurate.

In driver's recognition methods, device, equipment and medium provided by the embodiment of the present invention, by obtain user to It identifies behavioral data and target driving model, is based on behavioral data to be identified and target driving model, obtain identification probability value, lead to It crosses and judges whether identification probability value is greater than predetermined probabilities value and determines whether for my driving, so that driver's recognition result is more accurate Reliably.

Detailed description of the invention

In order to illustrate the technical solution of the embodiments of the present invention more clearly, below by institute in the description to the embodiment of the present invention Attached drawing to be used is needed to be briefly described, it should be apparent that, the accompanying drawings in the following description is only some implementations of the invention Example, for those of ordinary skill in the art, without any creative labor, can also be according to these attached drawings Obtain other attached drawings.

Fig. 1 is a flow chart of driving model training method in the embodiment of the present invention 1.

Fig. 2 is a specific flow chart of step S12 in Fig. 1.

Fig. 3 is a specific flow chart of step S121 in Fig. 2.

Fig. 4 is a specific flow chart of step S13 in Fig. 1.

Fig. 5 is a specific flow chart of step S14 in Fig. 1.

Fig. 6 is a functional block diagram of driving model training device in the embodiment of the present invention 2.

Fig. 7 is a flow chart of driver's recognition methods in the embodiment of the present invention 3.

Fig. 8 is a functional block diagram of driver's identification device in the embodiment of the present invention 4.

Fig. 9 is a schematic diagram of terminal device in the embodiment of the present invention 6.

Specific embodiment

Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete Site preparation description, it is clear that described embodiments are some of the embodiments of the present invention, instead of all the embodiments.Based on this hair Embodiment in bright, every other implementation obtained by those of ordinary skill in the art without making creative efforts Example, shall fall within the protection scope of the present invention.

Embodiment 1

Fig. 1 shows a flow chart of driving model training method in the present embodiment.The driving model training method can be applied On the terminal device of insurance institution or other mechanisms, to be identified using trained driving model, reach intelligent knowledge Other effect.It such as can be applicable on the terminal device of insurance institution, for training driving model corresponding with user, with convenience The user for handling vehicle insurance in insurance institution is identified with trained driving model, to determine whether to open for user Vehicle.As shown in Figure 1, the driving model training method includes the following steps:

S11: obtaining the training behavioral data of user, and training behavioral data is associated with user identifier.

Wherein, training behavioral data refers to the behavior number trained for carrying out driving model that user obtains in trip According to.Behavioral data including but not limited to refers to that any time collected speed, acceleration, angle and angle of the user in trip add At least one of data such as speed.User identifier is the mark for unique identification user, in order to guarantee to train what is obtained to drive Sail model can be used to identify whether be user drive, need to make all trained behavioral datas got with user identifier phase Association.Wherein, training behavioral data is associated with user identifier, refers to the training behavioral data that each user generates in trip Uniquely corresponding with user identifier, a user identifier can be associated multiple trained behavioral datas.

In the present embodiment, user in advance the application program on the mobile terminals such as mobile phone and plate (i.e. (Application, Abbreviation APP) on complete registration so that the corresponding server of application program can obtain corresponding user identifier.The user identifier can Think cell-phone number or identification card number of user etc. can unique identification user mark.When user carries mobile terminal trip, move Speed, acceleration, angle and the angle that built-in sensor can acquire any time during user goes on a journey in real time in dynamic terminal add The behavioral datas such as speed, can also in real time any time acquisition GPS positioning information, and based on GPS positioning information carry out calculate obtain Take corresponding behavioral data.After acquisition for mobile terminal to behavioral data, behavior data are uploaded onto the server, so that service The behavioral data that device will acquire is stored in the databases such as MySQL, Oracle, and makes each behavioral data and a user identifier Associated storage.When terminal device needs to carry out driving model training, acquisition can be inquired from the databases such as MySQL, Oracle Behavioral data associated with user identifier, the training behavioral data as training driving model.The training behavior number of the user According to a large amount of training datas comprising each user, enough training behavioral datas are capable of providing, are provided for driving model training Good data basis, to guarantee to train the recognition effect of obtained driving model.

Active user can be used in walking, bicycle, light cavalry, bus, car, railway and aircraft at least when going on a journey A kind of trip of mode of transportation, the behavioral datas such as the corresponding speed, acceleration of different modes of transportation, angle, angular acceleration are not identical. Therefore, the training behavioral data obtained in step S11 may be that the modes of transportation such as walking, bicycle, railway and aircraft are corresponding Behavioral data, there are larger differences for behavioral data when driving vehicle with user, if being directly based upon the training of step S11 acquisition Behavioral data carries out driving model training, may influence the recognition effect for the driving model that training obtains.

S12: based on training behavioral data, trained driving data associated with user identifier is obtained.

Wherein, training driving data refer to that user obtains when to drive a kind of trip of this mode of transportation of car for instructing Practice the behavioral data of driving model.It is to be appreciated that since each trained behavioral data is associated with user identifier, and training is driven Sailing data is one of trained behavioral data, so training driving data is associated with user identifier.Training driving data area The row acquired when walking, bicycle, railway, aircraft etc. are gone on a journey in a manner of car to drive is not used in training behavioral data For data, training driving data is obtained from training behavioral data, being conducive to the driving model that Support Training obtains can more preferably reflect The driving habit of user, to identify whether as user's driving.In the present embodiment, the training behavioral data of acquired original It can not be directly used in trained driving model, user need to be extracted in training behavioral data drive when car mode is gone on a journey and acquire Training driving data of the behavioral data as driving model.The training behavioral data of mobile terminal acquisition user is simultaneously stored in data In library, identification extracts in behavioral data and drives number as training with user's driving behavior data in various trained behavioral datas According to, so that the training driving data obtained can be applied to the training process of driving model, it is that the training process of driving model mentions For reliably training driving data.

As shown in Fig. 2, obtaining the training behavioral data of user in step S12, training behavioral data is related to user identifier Connection, specifically comprises the following steps:

S121: based on training behavioral data, behavior type corresponding with training behavioral data, behavior type and user are obtained Mark is associated.

Wherein, behavior type is user's trip mode of transportation corresponding with training behavioral data, the traffic of user's trip Mode may include the modes of transportation such as walking, bicycle, light cavalry, bus, car, railway and aircraft.Training behavioral data It may include the behavioral datas such as speed, acceleration, angle and angular acceleration.In the present embodiment, each behavior type all with it is corresponding User identifier it is associated, the application program on mobile terminal identifies trained behavior number according to the training behavioral data of acquisition According to the corresponding behavior type of middle difference training behavioral data, behavior type associated with user identifier is obtained.

Specifically, user A and user B can be used mobile terminal and upload behavioral data to server, so that terminal is set It is standby when carrying out driving model training, user A more a moment corresponding speed, acceleration can be obtained from database by server The training behavioral data such as degree, angle and angular acceleration obtains user B more a moment corresponding speed, acceleration, angle and angle and adds Speed etc. trains behavioral data, determines that the training behavioral data obtained belongs to user A or user B according to user identifier correlation, then Training behavioral data such as speed, acceleration, angle and angular acceleration etc. are handled, identify the training behavioral data of the user Corresponding behavior type is specifically to belong in the modes of transportation such as walking, bicycle, light cavalry, bus, car, railway and aircraft Any behavior type, to obtain and the corresponding behavior type of training behavioral data.

As shown in figure 3, based on training behavioral data, obtaining behavior class corresponding with training behavioral data in step S121 Type, behavior type is associated with user identifier, specifically comprises the following steps:

S1211: obtaining trained behavior type identification model, and behavior type identification model includes at least two cluster classes Cluster, the corresponding behavior type of each cluster class cluster, and each cluster class cluster includes a mass center.

Wherein, trained behavior type identification model is the corresponding behavior of trained behavioral data for identification in advance The model of type.Behavior type identification model is stored in advance in the database, when terminal device carries out driving model training, Behavior type identification model can be transferred from database.In the present embodiment, behavior type identification model is poly- by K-means Class algorithm carries out the model obtained after clustering processing to historical behavior data.The historical behavior data are user's acquisitions in trip For training the behavioral data of behavior type identification model, behavior data including but not limited to refer to times of the user when going on a journey At least one for the data such as moment collected speed, acceleration, angle and angular acceleration of anticipating.Wherein, K-means clustering algorithm It is a kind of clustering algorithm based on distance assessment similarity, i.e., the distance of two objects is closer, and the bigger cluster of similarity is calculated Method.

Specifically, the behavior type identification model obtained after being clustered using K-means clustering algorithm includes at least two A cluster class cluster, the corresponding behavior type of each cluster class cluster, and each cluster class cluster includes a mass center.In the present embodiment, It may include 7 cluster class clusters in the trained behavior type identification model, each cluster class cluster respectively represents walking, voluntarily Vehicle, light cavalry, bus, car, railway and aircraft, i.e., each cluster class cluster represent a kind of behavior type.Each trained behavior The centroid distance of data to cluster class cluster is smaller, then the training behavioral data is more likely to belong to the corresponding behavior of cluster class cluster Type.

S1212: the distance for training behavioral data to each mass center is calculated.

In the present embodiment, each cluster class cluster in the training behavioral data and at least two cluster class clusters of acquisition is calculated separately The distance of corresponding mass center, to determine the similitude of the training behavioral data and each cluster class cluster.By calculating training behavior The Euclidean distance of data and the corresponding mass center of each cluster class cluster, to evaluate training behavioral data according to the size of Euclidean distance With the similitude of each cluster class cluster.Euclidean distance (euclidean metric, also known as euclidean metric), which refers to, ties up sky in m Between in actual distance or vector between two points natural length (i.e. the distance of the point to origin).Any two n-dimensional vector The Euclidean distance of a (Xi1, Xi2 ..., Xin) and b (Xj1, Xj2 ..., Xjn)

S1213: will be apart from the corresponding behavior type of the smallest cluster class cluster, as the corresponding behavior class of training behavioral data Type.

It, will by calculating the Euclidean distance of training behavioral data and the corresponding mass center of each cluster class cluster in the present embodiment The corresponding behavior type of cluster class cluster belonging to the smallest mass center of the distance being calculated, as the corresponding row of training behavioral data For type.It is to be appreciated that training behavioral data with cluster closer at a distance from the corresponding behavior type of class cluster, the then training row The behavior type of cluster class cluster representative is more likely to belong to for data.Such as getting the speed of user A is 40km/s, is accelerated Degree is 5km/s2, and in behavior type identification model includes 7 cluster class clusters, then calculates separately the training behavioral data and 7 Cluster the Euclidean distance of the mass center of class cluster；Compare the size for calculating 7 Euclidean distances obtained again, by the smallest matter of Euclidean distance The corresponding behavior type of cluster class cluster belonging to the heart determines the purpose of the corresponding behavior type of training behavioral data.

S122: being the training behavioral data of driving style by behavior type, as training driving data.

Wherein, driving style refers to that one of behavior type corresponding with user identifier, in particular to user are going out The behavior type of drive manner trip is selected when row.In the present embodiment, terminal device identify with training behavioral data it is corresponding Behavior type after, choose wherein behavior type be driving style data, as training driving data, so as to utilize the training Driving data training for identification whether be user driving driving model.Specifically, terminal device is obtained from database The training behavioral data for taking family A may correspond to the behavior types such as walking, bus, car and aircraft, use step When S121 identifies training behavioral data, after determining the corresponding behavior type of each trained behavioral data, row is therefrom chosen Training driving data is used as the training behavioral data that type is driving style.Row is driven by choosing in a variety of behavior types It is available to carry out the required training driving data of driving model training for type.

S13: being based on user identifier, obtains positive negative sample from training driving data.

Wherein, user identifier refers to the mark for determining user identity, positive sample are to refer to identify to be to be identified User drive training driving data, negative sample be refer to identification be not to be identified user driving instruction Practice driving data.In the present embodiment, training driving data is extracted from training behavioral data, is trained behavioral data and is used Family mark is associated, therefore training driving data is also associated with user identifier, according to the user identifier of training driving data, obtains Take the positive negative sample for needing to carry out driving model training.

Since the behavior type of user is related to macroscopical road conditions, in the most of time of stroke, driving behavior is similar , do not have distinguishability, therefore trained driving data duration should be shortened, so that the training driving data obtained is more representative, And has higher distinguishability, and be conducive to save the training duration of driving model.In the present embodiment, step S13 is specifically wrapped It includes: based on user identifier, the data of preset data duration is chosen from training driving data as positive negative sample, are shortened with reaching The purpose of training driving data duration, so as to shorten the duration of driving model training.The preset data duration is that system is set in advance The duration acquired for limiting data set.Such as acquire data conduct in ten minutes when each run originates in trained driving data Positive negative sample, collected training drives when which can be driven out to cell just to drive car or just be driven out to ground library etc. Data.The required parameter in driving model can be effectively trained by the positive negative sample that training driving data obtains, is effectively prevent Training result is biased to extreme situation, so that the recognition effect of the driving model obtained by positive and negative sample training is more accurate.

As shown in figure 4, being based on user identifier in step S13, positive negative sample is obtained from training driving data, is specifically included Following steps:

S131: identifying in corresponding trained driving data from target user, chooses the corresponding training of preset time period and drives Data are as positive sample.

Wherein, target user refers to the driving model user to be identified.Correspondingly, target user's mark is for unique Identify the mark of target user.In the present embodiment, choose corresponding with target user's mark trained driving data, and will be default when Between the corresponding trained driving data of section as positive sample.Specifically, the positive sample can be target user A preset time period such as The training driving data of preceding 600s (i.e. preset data duration) in driving data is trained in continuous 2 months morning 8-9 points.In order to Further save driving model training duration, can make the corresponding trained driving data of positive sample be preset data duration every The data that one unit time obtained, as obtained primary training driving data every 10s in 600s before any trained driving data, 60 specific training driving datas can then be obtained as positive sample.

S132: it from the corresponding trained driving data of non-targeted user identifier, chooses the corresponding training of same period and drives Data are sailed as negative sample.

Wherein, non-targeted user refers to the other users other than the user that driving model to be identified.Correspondingly, non-targeted User identifier is the mark for the non-targeted user of unique identification.In the present embodiment, choose corresponding with non-targeted user identifier Training driving data, and using the corresponding trained driving data of preset time period as negative sample.It is to be appreciated that being selected in negative sample Take the corresponding preset time period of trained driving data identical as training driving data preset time period is chosen in positive sample, to guarantee Negative sample and positive sample are the training driving datas that not same user obtains under identical conditions.Specifically, which can be with It is that non-target user B or non-targeted user C are trained in driving data in preset time period such as continuous 2 months morning 8-9 points The training driving data of preceding 600s.In order to further save the training duration of driving model, the corresponding training of negative sample can be made to drive Sailing data is the data obtained in preset data duration every a unit time, the unit time of a unit time and positive sample Identical, 600s obtains primary training driving data every 10s before such as any trained driving data, obtains 60 specifically training altogether Driving data is as negative sample.

It further, is the accuracy for improving driving model training, in the corresponding driving model of training objective user, eventually End equipment also can receive the data query instruction of user's input, and data query instruction includes that target user identifies.Terminal device After receiving data query instruction, it is detailed that the corresponding target user of target user's mark is inquired by query sentence of database Information.Target user's details include the information such as home address, business address, the work hours of target user.Also, eventually End equipment further inquire in database with the presence or absence of with the same or similar non-targeted user of target user's details so that Terminal device can be inquired based on the corresponding non-targeted user identifier of non-targeted user and be obtained corresponding trained driving data conduct Negative sample, so that the details of positive negative sample are same or similar, so that collected target user and non-targeted user are corresponding Training driving data macroscopical road conditions it is substantially similar, be more advantageous in driving model training and guarantee the obtained driving mould of training The recognition accuracy of type.

S133: by the quantity of preset ratio configuration positive sample and negative sample.

Wherein, preset ratio refers to the ratio of initial pre-set positive sample and negative sample quantity.In the present embodiment, just The ratio of negative sample is mixed by 1:1, avoids over-fitting occur because the corresponding trained driving data quantity of positive negative sample is not identical Phenomenon.Wherein, over-fitting refers in order to obtain unanimously hypothesis and makes to assume to become over stringent phenomenon, and avoiding over-fitting is point A core missions in the design of class device.Specifically, there can be 60 specific training driving datas as positive sample and 60 Specific training driving data is as negative sample, wherein and the trained driving data of 60 of positive sample is collected in target user A, and 60 datas of negative sample can be collected between non-targeted user B, non-targeted user C or other non-targeted users arbitrarily to compare The ratio of 60 trained driving datas that example is composed, i.e., positive negative sample is mixed by 1:1.

S14: being trained error backward propagation method model using positive negative sample, obtains target driving model.

Wherein, error back propagation (Back Propagation, abbreviation BP) neural network model, before being a kind of multilayer Neural network is presented, communication process is signal propagated forward, error back propagation.Error backward propagation method model has Fault-tolerance and the strong advantage of generalization ability.Error backward propagation method model have input layer, output layer and hidden layer this Three layers of network structure, each layer include at least one neuron.Since positive negative sample respectively corresponds target user and non-targeted The training driving data of user is trained error backward propagation method model by using positive negative sample, may make The target driving model of acquisition may recognize that whether be that target user drives.Due to error backward propagation method model Have the advantages that fault-tolerance and generalization ability are strong, the recognition result of target driving model can be made more accurate.

As shown in figure 5, being instructed using the positive negative sample to error backward propagation method model in step S14 Practice, obtains target driving model, specifically comprise the following steps:

S141: initialization error back propagation artificial neural network model.

Wherein, error backward propagation method model include input layer, output layer and hidden layer model structure and Model parameter, the model parameter include the number of neuron in biasing and each layer between connection weight between layers, each layer Amount.It is to be appreciated that initialization error back propagation artificial neural network model, i.e. initialization error back propagation artificial neural network model Model parameter, connection weight between layers, in the biasing between each layer and each layer neuron quantity.

S142: inputting positive negative sample in error backward propagation method model, calculates the error back propagation mind Output valve through network model.

Specifically, positive negative sample is input in error backward propagation method model, is needed by calculating hidden layer Output valve so that acquire the output valve of output layer, i.e. the output valve of error backward propagation method model.Wherein, hidden layer Comprising an activation primitive, activation primitive is the function that weight results are converted to classification results, and effect is can be to nerve net Some non-linear factors are added in network, and neural network is allowed to better solve complex problem.

In step S142, positive negative sample is inputted in error backward propagation method model, calculates error back propagation The output valve of neural network model, specifically comprises the following steps:

S1421: it is calculated using positive negative sample of the activation primitive to input, obtains the output valve of hidden layer；Using sharp The calculation formula that function living calculates the positive negative sample of input includesWherein, H_jIt indicates The output valve of j-th of neuron of hidden layer；w_ijWeight of the expression input layer to hidden layer；x_iIndicate the input of input layer；Subscript n Indicate the quantity of input layer；Subscript i indicates i-th of neuron of input layer；Subscript j indicates j-th of neuron of hidden layer； a_jIndicate the biasing of input layer to hidden layer.

Wherein, select Sigmoid (S sigmoid growth curve) function as activation primitive, Sigmoid function is one in biology The function of common S type in, in information science, since singly properties, the Sigmoid function such as increasing and the increasing of inverse function list are normal for it It is used as the activation primitive of neural network, by variable mappings between 0-1.The function formula of Sigmoid function isThe output valve of hidden layer is calculated by the activation primitive of hidden layer.Specifically, using error back propagation mind Two classification problems, which are handled, through network model judges whether it is the problem of driving in person, for two classification problems, Sigmoid function By variable mappings to 0, between 1, functional value can be interpreted a belonging to the range of positive class (being to drive in person) probability just, therefore, Select Sigmoid letter as activation primitive, so that model training is more efficient.

S1422: calculating the output valve of hidden layer, obtains the output valve of output layer；To the output valve of hidden layer into Row calculate calculation formula includeWherein, k indicates k-th of neuron of output layer；Subscript j indicates hidden Hide j-th of neuron in layer；o_kIndicate the output of k-th of neuron of output layer；Subscript l indicates neuron in hidden layer Quantity；H_jIndicate the output valve of hidden layer；w_jkWeight of the expression hidden layer to output layer；b_kIndicate hidden layer to output layer Biasing.

Specifically, the output valve of neuron each in hidden layer is input to output layer to handle, by the output of acquisition Output valve of the output valve of layer as error backward propagation method model.

S143: error-duration model update is carried out to error backward propagation method model according to output valve, is obtained after updating Model parameter.

Specifically, the output valve to the error backward propagation method model got in step S142, i.e. step The output valve for the output layer that S142-12 is got carries out error-duration model update, i.e., suitable according to output layer-hidden layer-output layer Sequence is updated.

In step S143, error-duration model update is carried out to error backward propagation method model according to output valve, is obtained Updated model parameter, specifically comprises the following steps:

S1431: error-duration model update is carried out to error backward propagation method model according to output valve, is obtained after updating Model parameter；Model parameter includes the biasing between connection weight and each layer between layers；The meter of gradient descent algorithm Calculating formula includesAnd e_k=y_k-o_k, wherein o_kIt is exported for the prediction of model；y_kFor with o_kCorresponding expectation is defeated Out；e_kIndicate the error amount of k-th of neuron of output layer；Subscript m indicates output layer neuron quantity.

Specifically, using gradient descent algorithm to the model parameter, that is, layer and layer in error backward propagation method model Between connection weight and each layer between biasing be updated optimization.By the error back propagation got by step S142 mind The output valve of output layer through network model is updated in gradient descent algorithm calculation formula, to obtain updated model ginseng Number, gradient descent algorithm calculates simply, easy to accomplish.

S1432: it is based on formulaWith formula w_jk=w_jk+ηH_je_kTo model parameter In connection weight between layers be updated optimization；Wherein, η indicates learning rate.

Specifically, when being updated optimization with gradient descent algorithm, in the rule of right value update, it will usually in gradient Multiplied by a coefficient before, this coefficient is learning rate.Driving model can be made to instruct due to choosing inappropriate learning rate Experienced low efficiency, therefore, it is necessary to choose a suitable learning rate.In the present embodiment, the update optimization of model parameter is also wrapped The update to learning rate is included, the convergence rate of model is accelerated, improves the efficiency of model training.

In the present embodiment, in each iterative process, the update method of learning rate η can be pressed first by 3 multiple again 10 multiple gradually successively decreases to choose a series of η value, until finding one the smallest η value, while finding another maximum η Value.Maximum η value, or the η value more smaller than maximum value is required learning rate.

S1433: it is based on formulaWith formula b_k=b_k+ηe_kTo each in model parameter Biasing between layer is updated optimization；Wherein, a_jIndicate the biasing of input layer to hidden layer；b_kIndicate hidden layer to output layer Biasing.

Wherein, between each layer biasing include input layer to hidden layer biasing and hidden layer to output layer biasing.It can To understand ground, updated model parameter includes connection weight between each layer, the biasing between each layer and learning rate.

S143: being based on updated model parameter, obtains target driving model.

In the present embodiment, the updated model parameter that will acquire is applied in error backward propagation method model Target driving model can be obtained.Further, a probability value is eventually exported in the output layer of target driving model, the probability Value indicate information after being handled by target driving model with the target driving model close to degree, i.e. information input driving The probability of model have it is much, can be widely applied to driver identification, with reach accurately identify whether target user drive.

It is to be appreciated that after step S14, the driving model training method further include: the target driving model that will acquire Storage in the database, and creates model information table in the database, and model information table includes at least one model information, each Model information includes the storage address of user identifier and target driving model corresponding with user identifier in the database, so as to In when being identified using target driving model, corresponding target driving model can be inquired based on user identifier.

In driving model training method provided by the present embodiment, the training behavioral data of user, training behavior are first obtained Data are associated with user identifier, to be obtained respectively and target user's mark and non-targeted user identifier pair based on user identifier The training behavioral data answered, to guarantee that the target driving model of training acquisition can identify the driving behavior of target user.Then Based on training behavioral data, trained driving data associated with user identifier is obtained, which is never to go together To extract the corresponding trained behavioral data of driving style in type, the interference of other non-driving behavior data is excluded, is conducive to protect The recognition accuracy for the target driving model that card training obtains and the training effectiveness for improving target driving model, when saving training It is long, reliable, corresponding trained driving data is provided for the training process of driving model, to realize the training of driving model. Then it is based on user identifier, obtains positive negative sample from training driving data, positive negative sample can effectively determine that training objective drives Parameter needed for model guarantees the accuracy for the target driving model recognition result that training obtains.Finally use error back propagation Neural network model is trained, and is obtained target driving model, is trained using positive negative sample, it is anti-can effectively to update error Model parameter into Propagation Neural Network model, so that more smart by the driving model recognition effect that positive and negative sample training obtains It is quasi-.

It should be understood that the size of the serial number of each step is not meant that the order of the execution order in above-described embodiment, each process Execution sequence should be determined by its function and internal logic, the implementation process without coping with the embodiment of the present invention constitutes any limit It is fixed.

Embodiment 2

Fig. 6 shows the principle frame with the one-to-one driving model training device of driving model training method in embodiment 1 Figure.As shown in fig. 6, the driving model training device includes that training behavioral data obtains module 11, training driving data obtains mould Block 12, positive and negative sample acquisition module 13 and target driving model obtain module 14.Wherein, training behavioral data obtain module 11, Training driving data obtains module 12, positive and negative sample acquisition module 13 and target driving model obtain the realization function of module 14 with The corresponding step of driving model training method corresponds in embodiment 1, and to avoid repeating, the present embodiment is not described in detail one by one.

Training behavioral data obtains module 11, for obtaining the training behavioral data of user, training behavioral data and user Mark is associated.

Training driving data obtains module 12, for obtaining instruction associated with user identifier based on training behavioral data Practice driving data.

Preferably, it includes behavior type acquiring unit 121 that training driving data, which obtains module 12, and training driving data obtains Take unit 122.

Behavior type acquiring unit 121, for obtaining behavior corresponding with training behavioral data based on training behavioral data Type, behavior type are associated with user identifier.

Preferably, behavior type acquiring unit 121 includes that behavior type identification model obtains subelement 1211, distance calculates Subelement 1212 and behavior type determine subelement 1213.

Behavior type identification model obtains subelement 1211, for obtaining trained behavior type identification model, behavior Type identification model includes at least two cluster class clusters, the corresponding behavior type of each cluster class cluster, and each cluster class cluster packet Include a mass center.

Distance apart from computation subunit 1212, for calculating trained behavioral data to each mass center.

Behavior type determines subelement 1213, and being used for will be apart from the corresponding behavior type of the smallest cluster class cluster, as instruction Practice the corresponding behavior type of behavioral data.

Training driving data acquiring unit 122, for being the training behavioral data of driving style by behavior type, as instruction Practice driving data.

Positive and negative sample acquisition module 13 obtains positive negative sample from training driving data for being based on user identifier.

Preferably, positive and negative sample acquisition module 13 includes positive sample acquiring unit 131, negative sample acquiring unit 132 and ratio Example configuration unit 133.

Positive sample acquiring unit 131 chooses preset time for identifying in corresponding trained driving data from target user The corresponding trained driving data of section is as positive sample.

Negative sample acquiring unit 132, for choosing with for the moment from the corresponding trained driving data of non-targeted user identifier Between the corresponding trained driving data of section as negative sample.

Proportional arrangement unit 133, for the quantity by preset ratio configuration positive sample and negative sample.

Target driving model obtains module 14, for being trained using error backward propagation method model, obtains Target driving model.

Preferably, it includes network model initialization unit 141, output valve computing unit that target driving model, which obtains module 14, 142, right value update unit 143 and target driving model acquiring unit 144.

Network model initialization unit 141 is used for initialization error back propagation artificial neural network model.

Output valve computing unit 142 is calculated and is missed for inputting positive negative sample in error backward propagation method model The output valve of poor each layer of back propagation artificial neural network model.

Model parameter updating unit 143, for being carried out according to output valve to each layer of error backward propagation method model Error-duration model updates, and obtains updated model parameter.

Target driving model acquiring unit 144 obtains target driving model for being based on updated model parameter.

In driving model training device provided by the present embodiment, training behavioral data obtains module 11 for obtaining user Training behavioral data obtained respectively and mesh wherein training behavioral data associated with user identifier to be based on user identifier User identifier and the corresponding trained behavioral data of non-targeted user identifier are marked, it can with the target driving model for guaranteeing that training obtains Identify the driving behavior of target user.It trains driving data to obtain module 12 to be used for based on training behavioral data, acquisition and user Associated trained driving data is identified, which is never with the corresponding instruction of extraction driving style in behavior type Practice behavioral data, exclude the interference of other non-driving behavior data, advantageously ensures that the knowledge of the target driving model of training acquisition Other accuracy rate and the training effectiveness for improving target driving model are saved training duration, are provided for the training process of driving model Reliably, corresponding trained driving data, to realize the training of driving model.Positive and negative sample acquisition module 13 is used to be based on user Mark, from the positive negative sample of training driving data acquisition, parameter needed for positive negative sample effectively can determine training objective driving model, Guarantee the accuracy for the target driving model recognition result that training obtains.Target driving model obtains module 14 and is used for using positive and negative Sample is trained error backward propagation method model, obtains target driving model, by using positive negative sample to accidentally Poor back propagation artificial neural network model is trained, by the model parameter in the error backward propagation method model of initialization It is updated, so as to which the driving model with identification driver's function can be obtained, and can be realized through training acquisition Driving model recognition result more accurately effect.

Embodiment 3

Fig. 7 shows a flow chart of driver's recognition methods in the present embodiment.Driver's recognition methods can be applicable to guarantor On the terminal device of dangerous mechanism or other mechanisms, to identify to driver's driving behavior, reach the effect of intelligent recognition Fruit.As shown in fig. 7, driver's recognition methods includes the following steps:

S21: obtaining the behavioral data to be identified of user, and behavioral data to be identified is associated with user identifier.

Wherein, behavioral data to be identified refer to user trip when in real time it is collected whether be for identification target user The behavioral data that I drives.Behavioral data including but not limited to refer to user trip when any time collected speed, At least one of data such as acceleration, angle and angular acceleration.In the present embodiment, the behavioral data to be identified and user identifier Associated, the behavioral data to be identified for referring to that each user is formed in trip is associated with user identifier, to be based on the user The corresponding target driving model of identifier lookup is treated identification behavioral data and is identified.

S22: inquiring database based on user identifier, obtains target corresponding with user identifier and drives mould, wherein target Driving model is the model obtained using driving model training method in embodiment 1.

In the present embodiment, terminal device stores in the database according to the user identifier inquiry in behavioral data to be identified Target driving model, to identify whether behavioral data to be identified is the corresponding use of user identifier based on the target driving model Family drives.Wherein, target driving model and model information table are stored in database, model information table includes at least one Model information, each model information include user identifier and target driving model corresponding with user identifier in the database Storage address, in order to inquire corresponding target based on user identifier and drive when being identified using target driving model Sail model.Specifically, the behavioral data to be identified of user A can be obtained in real time for the mobile terminal of user A, and upload to service Device, so that the terminal device in insurance institution can obtain the behavioral data to be identified from server, and according to the row to be identified For the user identifier in data about user A, the target associated with user party A-subscriber's mark of inquiry storage in the database is driven The storage address of model is sailed, corresponding target driving model is obtained based on the storage address.

S23: being based on behavioral data to be identified and target driving model, obtains identification probability value.

In the present embodiment, behavioral data to be identified is input in target driving model and is identified, drives mould in target The conversion process based on each interlayer weight is carried out to the behavioral data to be identified of input in type, exports identification probability in output layer Value.Specifically, terminal device is after the behavioral data to be identified and target driving model for obtaining user A, by behavior number to be identified According to the conversion process based on each interlayer weight is carried out in target driving model, final identification probability value is obtained.The present embodiment In, which can real number between 0-1.

S24: judge whether identification probability value is greater than predetermined probabilities value；If identification probability value is greater than predetermined probabilities value, really It is set to me to drive.

Wherein, predetermined probabilities value is the pre-set probability value for evaluating whether to drive for me.In the present embodiment, Behavioral data to be identified is handled to the identification probability value finally obtained in target driving model, is compared with predetermined probabilities value Compared with.If identification probability value is greater than predetermined probabilities value, can be determined as driving in person.If identification probability value is less than or equal to default Probability value, then it is assumed that be not that I is driving.Specifically, if the identification probability value that terminal device obtains user A is 0.95, and it is pre- If probability value is 0.9, then it can be determined that user A drives.

In driver's recognition methods provided by the present embodiment, simultaneously based on the user identifier inquiry in behavioral data to be identified Corresponding target driving model is obtained, the acquisition process of target driving model is simple and fast.It is treated again using target driving model Identification behavioral data is identified, advantageous to ensure the accuracy for obtaining identification probability value.By judging whether identification probability value is big Determine whether to drive for me in predetermined probabilities value, i.e. determination is that the corresponding user of user identifier drives or user marks The car that corresponding user takes other users driving is known, to guarantee that driver's recognition result is more accurate reliable.

Embodiment 4

Fig. 8 shows the principle frame with the one-to-one driving model training device of driving model training method in embodiment 1 Figure.As shown in figure 8, the driving model training device includes that behavioral data to be identified obtains module 21, target driving model obtains Module 22, identification probability value obtain module 23 and recognition result judgment module 24.Wherein, behavioral data to be identified obtains module 21, target driving model obtains module 22, identification probability value obtains the realization function of module 23 and recognition result judgment module 24 Step corresponding with driving model training method in embodiment corresponds, and to avoid repeating, the present embodiment is not described in detail one by one.

Behavioral data to be identified obtains module 21, for obtaining the behavioral data to be identified of user, behavioral data to be identified It is associated with user identifier.

Target driving model obtains module 22, for inquiring database based on user identifier, obtains opposite with user identifier The target driving model answered.

Identification probability value obtains module 23, and for being based on behavioral data to be identified and target driving model, it is general to obtain identification Rate value.

Recognition result judgment module 24, for judging whether identification probability value is greater than predetermined probabilities value；If identification probability value Greater than predetermined probabilities value, it is determined that for my driving.

In driver's recognition methods device provided by the present embodiment, identification behavioral data is treated using target driving model It is identified, it is advantageous to ensure the accuracy for obtaining identification probability value.An identification probability value is exported in driving model output layer, is passed through Compared with predetermined probabilities value, the driver that may be implemented to treat identification behavioral data representative is effectively identified, to guarantee to drive It is more accurate reliable to sail people's recognition result.

Embodiment 5

The present embodiment provides a computer-readable medium, it is stored with computer program on the computer-readable medium, the meter Driving model training method in embodiment 1 is realized when calculation machine program is executed by processor, to avoid repeating, which is not described herein again. Alternatively, realizing the function of each module/unit of driving model training device in embodiment 2 when the computer program is executed by processor Can, to avoid repeating, which is not described herein again.Alternatively, realizing driver in embodiment 3 when the computer program is executed by processor The function of each step does not repeat one by one herein in recognition methods to avoid repeating.Alternatively, the computer program is held by processor Realize that the function of each module/unit in driver's identification device in embodiment 4 does not repeat one by one herein to avoid repeating when row.

Embodiment 6

Fig. 9 is a schematic diagram of the terminal device that one embodiment of the invention provides.As shown in figure 9, the terminal of the embodiment Equipment 90 includes: processor 91, memory 92 and is stored in the computer that can be run in memory 92 and on processor 91 Program 93 realizes the driving model training method in embodiment 1 when the computer program is executed by processor 91, to avoid weight It is multiple, it does not repeat one by one herein.Alternatively, realizing driving model training in embodiment 2 when the computer program is executed by processor 91 The function of each model/unit does not repeat one by one herein in device to avoid repeating.Alternatively, the computer program is by processor 91 Realize that the function of each step in driver's recognition methods in embodiment 3 does not repeat one by one herein to avoid repeating when execution.Or Person realizes the function of each module/unit in driver's identification device in embodiment 4 when the computer program is executed by processor 91 Energy.To avoid repeating, do not repeat one by one herein.

Illustratively, computer program 93 can be divided into one or more module/units, one or more mould Block/unit is stored in memory 92, and is executed by processor 91, to complete the present invention.One or more module/units can To be the series of computation machine program instruction section that can complete specific function, the instruction segment is for describing computer program 93 at end Implementation procedure in end equipment 90.For example, computer program 90 can be divided into the acquisition of the training behavioral data in embodiment 2 Module 11, training driving data obtain module 12, positive and negative sample acquisition module 13 and target driving model and obtain module 14, or Behavioral data to be identified in embodiment 4 obtains module 21, target driving model obtains module 22, identification probability value obtains module 23 and recognition result judgment module 24, the concrete function of each module will not repeat them here as described in embodiment 2 or embodiment 4.

Terminal device 90 can be desktop PC, notebook, palm PC and cloud server etc. and calculate equipment.Eventually End equipment may include, but be not limited only to, processor 91, memory 92.It will be understood by those skilled in the art that Fig. 9 is only eventually The example of end equipment 90 does not constitute the restriction to terminal device 90, may include components more more or fewer than diagram, or Combine certain components or different components, for example, terminal device can also include input-output equipment, network access equipment, Bus etc..

Alleged processor 91 can be central processing unit (Central Processing Unit, CPU), can also be Other general processors, digital signal processor (Digital Signal Processor, DSP), specific integrated circuit (Application Specific Integrated Circuit, ASIC), ready-made programmable gate array (Field- Programmable Gate Array, FPGA) either other programmable logic device, discrete gate or transistor logic, Discrete hardware components etc..General processor can be microprocessor or the processor is also possible to any conventional processor Deng.

Memory 92 can be the internal storage unit of terminal device 90, such as the hard disk or memory of terminal device 90.It deposits Reservoir 92 is also possible to the plug-in type hard disk being equipped on the External memory equipment of terminal device 90, such as terminal device 90, intelligence Storage card (Smart Media Card, SMC), secure digital (Secure Digital, SD) card, flash card (Flash Card) Deng.Further, memory 92 can also both including terminal device 90 internal storage unit and also including External memory equipment.It deposits Reservoir 92 is for storing other programs and data needed for computer program and terminal device.Memory 92 can be also used for temporarily When store the data that has exported or will export.

It is apparent to those skilled in the art that for convenience of description and succinctly, only with above-mentioned each function Can unit, module division progress for example, in practical application, can according to need and by above-mentioned function distribution by different Functional unit, module are completed, i.e., the internal structure of described device is divided into different functional unit or module, more than completing The all or part of function of description.Each functional unit in embodiment, module can integrate in one processing unit, can also To be that each unit physically exists alone, can also be integrated in one unit with two or more units, it is above-mentioned integrated Unit both can take the form of hardware realization, can also realize in the form of software functional units.In addition, each function list Member, the specific name of module are also only for convenience of distinguishing each other, the protection scope being not intended to limit this application.Above system The specific work process of middle unit, module, can refer to corresponding processes in the foregoing method embodiment, and details are not described herein.

In the above-described embodiments, it all emphasizes particularly on different fields to the description of each embodiment, is not described in detail or remembers in some embodiment The part of load may refer to the associated description of other embodiments.

Those of ordinary skill in the art may be aware that list described in conjunction with the examples disclosed in the embodiments of the present disclosure Member and algorithm steps can be realized with the combination of electronic hardware or computer software and electronic hardware.These functions are actually It is implemented in hardware or software, the specific application and design constraint depending on technical solution.Professional technician Each specific application can be used different methods to achieve the described function, but this realization is it is not considered that exceed The scope of the present invention.

In embodiment provided by the present invention, it should be understood that disclosed device/terminal device and method, it can be with It realizes by another way.For example, device described above/terminal device embodiment is only schematical, for example, institute The division of module or unit is stated, only a kind of logical function partition, there may be another division manner in actual implementation, such as Multiple units or components can be combined or can be integrated into another system, or some features can be ignored or not executed.Separately A bit, shown or discussed mutual coupling or direct-coupling or communication connection can be through some interfaces, device Or the INDIRECT COUPLING or communication connection of unit, it can be electrical property, mechanical or other forms.

The unit as illustrated by the separation member may or may not be physically separated, aobvious as unit The component shown may or may not be physical unit, it can and it is in one place, or may be distributed over multiple In network unit.It can select some or all of unit therein according to the actual needs to realize the mesh of this embodiment scheme 's.

It, can also be in addition, the functional units in various embodiments of the present invention may be integrated into one processing unit It is that each unit physically exists alone, can also be integrated in one unit with two or more units.Above-mentioned integrated list Member both can take the form of hardware realization, can also realize in the form of software functional units.

If the integrated module/unit be realized in the form of SFU software functional unit and as independent product sale or In use, can store in a computer-readable medium.Based on this understanding, the present invention realizes above-described embodiment side All or part of the process in method can also instruct relevant hardware to complete, the computer by computer program Program can be stored in a computer-readable medium, and the computer program is when being executed by processor, it can be achieved that above-mentioned each side The step of method embodiment.Wherein, the computer program includes computer program code, and the computer program code can be Source code form, object identification code form, executable file or certain intermediate forms etc..The computer-readable medium can wrap It includes: any entity or device, recording medium, USB flash disk, mobile hard disk, magnetic disk, light of the computer program code can be carried Disk, computer storage, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium etc..It should be noted that the computer The content that readable medium includes can carry out increase and decrease appropriate according to the requirement made laws in jurisdiction with patent practice, such as In certain jurisdictions, according to legislation and patent practice, computer-readable medium do not include be electric carrier signal and telecommunications letter Number.

Embodiment described above is merely illustrative of the technical solution of the present invention, rather than its limitations；Although referring to aforementioned reality Applying example, invention is explained in detail, those skilled in the art should understand that: it still can be to aforementioned each Technical solution documented by embodiment is modified or equivalent replacement of some of the technical features；And these are modified Or replacement, the spirit and scope for technical solution of various embodiments of the present invention that it does not separate the essence of the corresponding technical solution should all It is included within protection scope of the present invention.

Claims

1. a kind of driving model training method characterized by comprising

Error backward propagation method model is trained using the positive negative sample, obtains target driving model；

Wherein, described to be based on the user identifier, positive negative sample is obtained from the trained driving data, comprising:

Data query instruction is received, the data query instruction includes that target user identifies；

It is identified in the corresponding trained driving data from the target user, chooses the corresponding training of preset time period and drive Data are sailed as positive sample；

Inquiry database is identified based on the target user, obtains target user's details；Target user's details Including home address, business address and work hours；

Database is inquired based on target user's details, is obtained the same or similar with target user's details At least one non-targeted user, the corresponding non-targeted user identifier of non-targeted user；

From the corresponding trained driving data of the non-targeted user identifier, the corresponding training of same period is chosen Driving data is as negative sample；

The quantity of the positive sample and the negative sample is configured by preset ratio.

2. driving model training method as described in claim 1, which is characterized in that it is described to be based on the trained behavioral data, Obtain trained driving data associated with the user identifier, comprising:

Based on the trained behavioral data, obtain behavior type corresponding with the trained behavioral data, the behavior type and The user identifier is associated；

It is the trained behavioral data of driving style by the behavior type, as the trained driving data.

3. driving model training method as described in claim 1, which is characterized in that described to use the positive negative sample to error Back propagation artificial neural network model is trained, and obtains target driving model, comprising:

Initialize the error backward propagation method model；

The positive negative sample is inputted in the error backward propagation method model, calculates the error back propagation nerve The output valve of network model；

Error-duration model update is carried out to the error backward propagation method model according to the output valve, is obtained updated Model parameter；

Based on the updated model parameter, target driving model is obtained.

4. driving model training method as claimed in claim 3, which is characterized in that described in the error back propagation nerve The positive negative sample is inputted in network model, calculates the output valve of the error backward propagation method model, comprising:

It is calculated using the positive negative sample of the activation primitive to input, obtains the output valve of hidden layer；It is described to use activation The calculation formula that function calculates the positive negative sample of input includesWherein, H_jIndicate institute State the output valve of j-th of neuron of hidden layer；w_ijWeight of the expression input layer to hidden layer；x_iIndicate the defeated of the input layer Enter；Subscript n indicates the quantity of the input layer；Subscript i indicates described i-th of neuron of input layer；Subscript j indicates institute State j-th of neuron of hidden layer；a_jIndicate the biasing of the input layer to the hidden layer；

The output valve of the hidden layer is calculated, the output valve of output layer is obtained；The output valve to the hidden layer The calculation formula calculated includesWherein, k indicates described k-th of neuron of output layer；Subscript j Indicate j-th of neuron in the hidden layer；o_kIndicate the output of k-th of neuron of the output layer；Subscript l indicates institute State the quantity of neuron in hidden layer；H_jIndicate the output valve of the hidden layer；w_jkIndicate the hidden layer to the output layer Weight；b_kIndicate the biasing of the hidden layer to the output layer；

It is described that error-duration model update is carried out to the error backward propagation method model according to the output valve, it obtains and updates Model parameter afterwards, comprising:

Optimization is updated to the model parameter in error backward propagation method model using gradient descent algorithm, is obtained more Model parameter after new；The model parameter includes the biasing between connection weight and each layer between layers；The gradient The calculation formula of descent algorithm includesAnd e_k=y_k-o_k, wherein o_kIt is exported for the prediction of model；y_kFor with the o_k Corresponding desired output；e_kIndicate the error amount of k-th of neuron of the output layer；Subscript m indicates the output layer nerve First quantity；

Based on formulaWith formula w_jk=w_jk+ηH_je_kTo the institute in the model parameter The connection weight stated between layers is updated optimization；Wherein, η indicates learning rate；

Based on formulaWith formula b_k=b_k+ηe_kTo each layer in the model parameter Between biasing be updated optimization；Wherein, a_jIndicate the biasing of input layer to hidden layer；b_kIndicate hidden layer to output layer Biasing.

5. a kind of driver's recognition methods, comprising:

Database is inquired based on the user identifier, obtains target driving model corresponding with the user identifier, the mesh Mark driving model is the model obtained using any one of the claim 1-4 driving model training method；

Judge whether the identification probability value is greater than predetermined probabilities value；If the identification probability value is greater than the predetermined probabilities value, Then it is determined as driving in person.

6. a kind of driving model training device characterized by comprising

Training behavioral data obtains module, and for obtaining the training behavioral data of user, the trained behavioral data and user are marked Sensible association；Each trained behavioral data and user identifier associated storage；

Training driving data obtains module, for being based on the trained behavioral data, obtains associated with the user identifier Training driving data；

Target driving model obtains module, is trained using the positive negative sample to error backward propagation method model, Obtain target driving model；

Wherein, the positive and negative sample acquisition module includes:

Data query instruction receiving unit, inquiry instruction, the data query instruction include that target user marks for receiving data Know；

Positive sample acquiring unit chooses preset time period for identifying in the corresponding trained driving data from target user The corresponding trained driving data is as positive sample；

Target user's details acquiring unit obtains the target user and identifies corresponding target user's details；It is described Target user's details include home address, business address and work hours；

Non-targeted user's acquiring unit is inquired database based on target user's details, is obtained and the target user At least one the same or similar non-targeted user of details, the corresponding non-targeted user identifier of non-targeted user；

Negative sample acquiring unit, for choosing same from the corresponding trained driving data of the non-targeted user identifier Period, the corresponding trained driving data was as negative sample；

Proportional arrangement unit, for configuring the quantity of the positive sample and the negative sample by preset ratio.

7. a kind of driver's identification device characterized by comprising

Behavioral data to be identified obtains module, for obtaining the behavioral data to be identified of user, the behavioral data to be identified with User identifier is associated；

Target driving model obtains module, for inquiring database based on the user identifier, obtains and the user identifier phase Corresponding target driving model；Wherein, the target driving model is using any one of the claim 1-4 driving model instruction Practice the model that method obtains；

8. a kind of terminal device, including memory, processor and storage are in the memory and can be on the processor The computer program of operation, which is characterized in that the processor realizes such as Claims 1-4 when executing the computer program The step of any one driving model training method；Alternatively, the processor is realized when executing the computer program as weighed Benefit requires the step of 5 driver's recognition methods.

9. a kind of computer readable storage medium, the computer-readable recording medium storage has computer program, and feature exists In realizing the driving model training method as described in any one of Claims 1-4 when the computer program is executed by processor Step；Alternatively, the processor realizes the step of driver's recognition methods as claimed in claim 5 when executing the computer program Suddenly.