A method, device, computer equipment, and storage medium for identifying the risk of drunk driving.

Abstract

This application belongs to the field of data analysis and relates to a method for identifying drunk driving risks. The method includes obtaining drunk driving characteristic data and corresponding drunk driving voice data from drunk driving cases; obtaining basic feature parameters based on the drunk driving characteristic data; obtaining voice feature parameters based on the drunk driving voice data; constructing a feature parameter set by combining the basic feature parameters and the voice feature parameters, and using the feature parameter set as model parameters to build an initial drunk driving risk identification model; training the initial drunk driving risk identification model to obtain a target drunk driving risk identification model; inputting the target drunk driving characteristics into the target drunk driving risk identification model, and outputting a risk prediction result. This application also provides a drunk driving risk identification device, computer equipment, and storage medium. Furthermore, this application relates to blockchain technology, and the feature parameter set can be stored in the blockchain. This application can improve the efficiency and accuracy of drunk driving risk identification.

Description

Technical Field

[0001] This application relates to the field of data analysis technology, and in particular to a method, device, computer equipment and storage medium for identifying drunk driving risks. Background Technology

[0002] Drunk driving risk identification refers to the systematic and continuous process of recognizing the risks of drunk driving and analyzing the potential causes of traffic accidents using various methods after they occur. The drunk driving risk identification process includes two stages: risk perception and risk analysis.

[0003] Traditional drunk driving risk identification solutions primarily employ common scenario analysis methods to uncover drunk driving risk characteristics and then identify them using structured rules. However, this approach suffers from low identification efficiency and a high false trigger rate. Furthermore, in vehicle damage compensation cases, drunk drivers often exhibit specific behavioral patterns, and traditional drunk driving risk identification solutions show low accuracy in predicting user drunk driving behavior. Summary of the Invention

[0004] The purpose of this application is to provide a method, device, computer equipment, and storage medium for identifying the risk of drunk driving, in order to solve the problems in related technologies.

[0005] To address the aforementioned technical problems, this application provides a method for identifying the risk of drunk driving, employing the following technical solution:

[0006] Obtain drunk driving cases, and extract drunk driving characteristic data and corresponding drunk driving voice data from the drunk driving cases;

[0007] Basic features are extracted from the drunk driving feature data, and the initial basic weights of the basic features are determined according to the drunk driving cases. Based on the basic features and the corresponding initial basic weights, basic feature parameters are obtained.

[0008] Based on the drunk driving voice data, voice features are obtained, and the initial voice weights of the voice features are determined. Based on the voice features and the corresponding initial voice weights, voice feature parameters are obtained.

[0009] The basic feature parameters and the voice feature parameters are combined to form a feature parameter set, and the feature parameter set is used as model parameters to construct an initial drunk driving risk identification model;

[0010] The initial drunk driving risk identification model is trained to obtain the target drunk driving risk identification model;

[0011] Obtain the target drunk driving characteristics, input the target drunk driving characteristics into the target drunk driving risk identification model, and output the risk prediction result.

[0012] Furthermore, the step of determining the initial basic weights of the basic features based on the drunk driving case includes:

[0013] The number of drunk driving cases corresponding to each feature category under the basic features is counted to obtain the number of categories;

[0014] Calculate the percentage of each feature category relative to the total number of drunk driving cases;

[0015] The initial basic weight of the basic feature under each feature category is obtained based on the percentage of drunk driving.

[0016] Furthermore, the step of obtaining voice features based on the drunk driving voice data includes:

[0017] Based on the acoustic features in the drunk driving voice data, the drunk driving voice data is processed by voice segmentation to obtain voice segments;

[0018] The speech segment is processed by speech recognition to obtain the speech recognition result corresponding to the drunk driving speech data;

[0019] Speech features are obtained based on the speech recognition results.

[0020] Furthermore, the step of determining the initial speech weights of the speech features includes:

[0021] Count the total number of occurrences of all the speech features and the number of features for each speech feature;

[0022] The initial speech weights of the speech features are calculated based on the number of features and the total number of occurrences.

[0023] Furthermore, the step of training the initial drunk driving risk identification model to obtain the target drunk driving risk identification model includes:

[0024] A training dataset and a validation dataset are obtained based on the drunk driving feature data and the drunk driving voice data;

[0025] The model parameters of the initial drunk driving risk identification model are adjusted based on the training dataset until the model converges, thus obtaining the model to be verified.

[0026] The verification dataset is input into the model to be verified for verification to obtain the verification result. When the verification result is greater than or equal to a preset threshold, the model to be verified is determined to be the target drunk driving risk identification model.

[0027] Furthermore, the step of adjusting the model parameters of the initial drunk driving risk identification model based on the training dataset includes:

[0028] The training dataset is input into the initial drunk driving risk identification model to obtain the drunk driving result;

[0029] Determine the feature contribution of each feature parameter in the model parameters based on the drunk driving results;

[0030] The adjustment coefficient of the corresponding feature parameter is determined based on the feature contribution degree;

[0031] The model parameters are adjusted according to the adjustment coefficients.

[0032] Furthermore, the step of inputting the verification dataset into the model to be verified for verification, obtaining the verification result, and determining the model to be verified as the target drunk driving risk identification model when the verification result is greater than or equal to a preset threshold includes:

[0033] Input the validation dataset into the model to be validated to obtain the prediction output results;

[0034] Calculate the prediction accuracy based on the predicted output results;

[0035] If the prediction accuracy is greater than or equal to a preset threshold, the model to be verified is output as the target drunk driving risk identification model.

[0036] If the prediction accuracy is less than a preset threshold, then the training dataset is updated, and the step of adjusting the model parameters of the initial drunk driving risk identification model based on the training dataset is executed.

[0037] To address the aforementioned technical problems, this application also provides a drunk driving risk identification device, which employs the following technical solution:

[0038] The acquisition module is used to acquire drunk driving cases and to acquire drunk driving feature data and corresponding drunk driving voice data from the drunk driving cases.

[0039] The basic feature extraction module is used to extract basic features from the drunk driving feature data, determine the initial basic weights of the basic features based on the drunk driving cases, and obtain basic feature parameters based on the basic features and the corresponding initial basic weights.

[0040] The speech feature extraction module is used to obtain speech features based on the drunk driving speech data, determine the initial speech weight of the speech features, and obtain speech feature parameters based on the speech features and the corresponding initial speech weights.

[0041] The construction module is used to combine the basic feature parameters and the voice feature parameters into a feature parameter set, and use the feature parameter set as model parameters to construct an initial drunk driving risk recognition model;

[0042] The training module is used to train the initial drunk driving risk identification model to obtain the target drunk driving risk identification model;

[0043] The prediction module is used to acquire the characteristics of the target drunk driving, input the characteristics of the target drunk driving into the target drunk driving risk identification model, and output the risk prediction result.

[0044] To address the aforementioned technical problems, this application also provides a computer device that employs the following technical solution:

[0045] The computer device includes a memory and a processor, the memory storing computer-readable instructions, and the processor executing the computer-readable instructions to implement the steps of the drunk driving risk identification method as described above.

[0046] To address the aforementioned technical problems, this application also provides a computer-readable storage medium, employing the technical solution described below:

[0047] The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of the drunk driving risk identification method as described above.

[0048] Compared with the prior art, the embodiments of this application have the following main advantages:

[0049] This application obtains drunk driving case data, including drunk driving feature data and corresponding drunk driving voice data. It extracts basic features from the drunk driving feature data, determines initial basic weights for these features based on the drunk driving cases, and obtains basic feature parameters based on the basic features and their corresponding initial basic weights. It then obtains voice features from the drunk driving voice data, determines initial voice weights for these voice features, and obtains voice feature parameters based on the voice features and their corresponding initial voice weights. The application combines the basic feature parameters and voice feature parameters to form a feature parameter set, which is used as model parameters to construct an initial drunk driving risk identification model. This initial drunk driving risk identification model is trained to obtain a target drunk driving risk identification model. The target drunk driving features are then obtained, input into the target drunk driving risk identification model, and the risk prediction result is output. This application constructs a drunk driving risk identification model using basic feature parameters and voice feature parameters, achieving a structured, multi-factor combined approach to risk identification. This eliminates unstable factors caused by the experience of risk control personnel, reducing the false trigger rate. Furthermore, by adjusting the parameters during the training of the drunk driving identification model, the sensitivity of the model can be flexibly controlled, improving the efficiency and accuracy of drunk driving risk identification. Attached Figure Description

[0050] To more clearly illustrate the solutions in this application, the accompanying drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 This is an exemplary system architecture diagram to which this application can be applied;

[0052] Figure 2 This is a flowchart of an embodiment of the drunk driving risk identification method according to this application;

[0053] Figure 3 This is a schematic diagram of one embodiment of the drunk driving risk identification device according to this application;

[0054] Figure 4 This is a schematic diagram of the structure of one embodiment of the computer device according to this application. Detailed Implementation

[0055] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein in the specification of the application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application; the terms "comprising" and "having," and any variations thereof, in the specification, claims, and foregoing drawings of this application, are intended to cover non-exclusive inclusion. The terms "first," "second," etc., in the specification, claims, or foregoing drawings of this application are used to distinguish different objects, not to describe a particular order.

[0056] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0057] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0058] This application provides a method for identifying the risk of drunk driving, which can be applied to, for example... Figure 1In the system architecture 100 shown, the system architecture 100 may include terminal devices 101, 102, and 103, a network 104, and a server 105. The network 104 is used as a medium to provide a communication link between the terminal devices 101, 102, and 103 and the server 105. The network 104 may include various connection types, such as wired or wireless communication links or fiber optic cables, etc.

[0059] Users can use terminal devices 101, 102, and 103 to interact with server 105 via network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, and 103, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social media platform software, etc.

[0060] Terminal devices 101, 102, and 103 can be various electronic devices with displays and support web browsing, including but not limited to smartphones, tablets, e-book readers, MP3 players (Moving Picture Experts Group Audio Layer III), MP4 players (Moving Picture Experts Group Audio Layer IV), laptops, and desktop computers, etc.

[0061] Server 105 can be a server that provides various services, such as a backend server that supports the pages displayed on terminal devices 101, 102, and 103.

[0062] It should be noted that the drunk driving risk identification method provided in this application embodiment is generally executed by a server / terminal device, and correspondingly, the drunk driving risk identification device is generally installed in the server / terminal device.

[0063] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0064] Continue to refer to Figure 2 The flowchart illustrates an embodiment of the drunk driving risk identification method according to this application, including the following steps:

[0065] Step S201: Obtain drunk driving cases, and extract drunk driving feature data and corresponding drunk driving voice data from the drunk driving cases.

[0066] In this embodiment, a request for drunk driving behavior analysis is received, and the corresponding drunk driving case is retrieved from the corresponding database. The drunk driving case includes drunk driving characteristic data and drunk driving voice data corresponding to the case. The drunk driving characteristic data includes, but is not limited to, case information, activity trajectory data, and user-related information. Case information includes the reporting time, incident time, incident location, reporting interval, and incident cause, etc. Activity trajectory data includes the activity range, activity location, activity time, number of times and time of the user's appearance in entertainment venues, and number of times and time of the user's appearance in catering venues, etc. User-related information includes the driver's gender and age listed on the policy, the driver who reported the incident, and the insured, etc.

[0067] The drunk driving voice data includes the reporting voice calls corresponding to drunk driving cases, and the recording data of drunk driving situations.

[0068] In this embodiment, upon receiving a drunk driving behavior analysis request, the request can be parsed to obtain parsed data. This parsed data can then be validated. If the validation passes, drunk driving cases can be extracted from the corresponding database according to preset rules. Alternatively, based on the received drunk driving behavior analysis request, eligible drunk driving cases can be directly retrieved from the corresponding database. The preset rules could be to retrieve drunk driving cases within a preset time period, but are not limited to this.

[0069] Step S202: Extract basic features from drunk driving feature data, determine the initial basic weights of the basic features based on drunk driving cases, and obtain basic feature parameters based on the basic features and the corresponding initial basic weights.

[0070] In this embodiment, the drunk driving characteristic data is analyzed to obtain basic drunk driving characteristics. These basic drunk driving characteristics include, but are not limited to, the time of the report, the time of the incident, the location of the incident, the time interval between reports, the cause of the incident, the gender and age of the driver listed on the policy, the driver who reported the incident, the insured, and activity trajectory data, etc. Specifically, keywords are extracted from the drunk driving characteristic data, and basic characteristics are determined based on the attributes of the keywords. For example, if a car collision occurred in the suburbs at 9 pm on April 30, 2021, due to a lane change, and the report was received at 9:30 pm that same day, the extracted keywords would be: April 30, 2021, 9 pm, 9:30 pm that same day, suburbs, lane change, etc. The corresponding basic characteristics obtained based on the attributes of the keywords would be: the date of the incident, the time of the incident, the time of the report, the location of the incident, the cause of the incident, etc.

[0071] In this embodiment, before extracting basic features based on drunk driving feature data, the drunk driving feature data is cleaned by sequentially performing data deduplication, deletion of abnormal data, cleaning of null values, and standardization of data format to obtain cleaned drunk driving feature data.

[0072] In some alternative implementations, the steps for determining the initial basic weights of basic features based on drunk driving cases include:

[0073] The number of drunk driving cases corresponding to each feature category under the basic features is counted to obtain the number of categories.

[0074] Calculate the percentage of each feature category relative to the total number of drunk driving cases;

[0075] The initial basic weights of the basic features under each feature category are obtained based on the proportion of drunk driving.

[0076] In this embodiment, for each basic feature, the drunk driving feature data is classified according to preset dimensions, including time dimension, spatial dimension (including region dimension, location dimension, etc.), age dimension, etc.

[0077] For example, when the basic feature is the time of the incident, the drunk driving feature data can be classified according to the month dimension, time dimension, and region dimension. The month dimension includes January to December, the time dimension can be 0-5 o'clock (where 5 o'clock includes 5:00-5:59, and the same applies below), 6-11 o'clock, 12-14 o'clock, 15-18 o'clock, 19-20 o'clock, and 21-23 o'clock, and the region dimension is divided according to the latitude and longitude of the administrative region, including the North Region, East Region, South Region, and West Region. Then, the initial basic weight of the basic feature time of the incident under each feature category is calculated.

[0078] Specifically, extract all drunk driving cases tagged with "drunk driving" from a preset year, such as 2010 to 2021. Divide the drunk driving cases by region to obtain drunk driving cases in different regions. Count the number of drunk driving cases in each region for each month and calculate the proportion of drunk driving cases in each region for each month (i.e., the drunk driving percentage). Scale up the drunk driving percentage proportionally and round it to obtain the weight of the month relative to drunk driving. Use the same method to determine the weight of time relative to drunk driving. For example, the initial basic weight of the North region in January is 3, the initial basic weight of the South region in January is 2, the initial basic weight of the East region in January is 2, and the initial basic weight of the West region in January is 3.

[0079] In this embodiment, when the basic feature is the location of the incident, it is classified according to different location tags. The location tags include entertainment venues, catering establishments, urban areas, suburbs, highways, and rural areas. Entertainment venues include clubs, KTVs, bars, etc., and catering establishments include restaurants, hotels, food stalls, seafood streets, etc.

[0080] Specifically, all drunk driving cases containing the "drunk driving" tag from 2010 to 2021 were extracted. After removing the administrative division names (including provinces, cities, districts, counties, townships, towns, villages, and streets) from the locations of the drunk driving incidents, the remaining keywords were used as location tags. The number of cases corresponding to each location tag was counted, and the proportion of drunk driving cases for each location tag was calculated (i.e., the drunk driving percentage). The drunk driving percentage was proportionally amplified and rounded to obtain the weight of the location tag relative to drunk driving.

[0081] In this embodiment, the basic features also include the reporting time interval, the cause of the accident (including lane change, turning, going straight, damage to the vehicle, and damage to the vehicle), the gender of the driver listed on the policy, the age of the driver listed on the policy, and whether the driver who reported the accident and the insured are the same person, etc. The corresponding initial basic weights can be calculated according to the method for determining the time and location of the accident as described above, which will not be elaborated here.

[0082] In some alternative implementations, the initial base weights of the basic features can be determined by a trained weight model. The weight model can include various types, such as a Long Short-Term Memory (LSTM) network + Conditional Random Field (CRF) model, or a Gradient Boosting Decision Tree (GBDT) model, etc.

[0083] Specifically, a basic feature training dataset carrying actual score labels is obtained, and the training dataset is input into a pre-constructed initial weight model to output predicted weight values. A loss function is calculated based on the predicted weight values, and the model parameters are adjusted according to the loss function. Iterative training continues until the model converges, and the final weight model is output. The basic weights of the basic features in the drunk driving feature data are then calculated using this weight model.

[0084] The basic features and the corresponding calculated initial basic weights are used as the basic feature parameters for constructing the initial drunk driving risk identification model.

[0085] Step S203: Obtain speech features based on drunk driving speech data, determine the initial speech weights of the speech features, and obtain speech feature parameters based on the speech features and the corresponding initial speech weights.

[0086] In this embodiment, the drunk driving voice data is subjected to speech recognition and converted into a text-based speech recognition result. Speech features are obtained from the speech recognition result. The speech features are speech keywords related to drunk driving, including drinking, alcohol, hotel, alcohol, eating, drinking, singing, restaurant, etc.

[0087] The steps for determining the initial speech weights of the speech features mentioned above include:

[0088] Calculate the total number of occurrences of all speech features and the number of features for each speech feature;

[0089] The initial speech weights of the speech features are calculated based on the number of features and the total number of occurrences.

[0090] Specifically, based on the number of features and the total number of occurrences, the proportion of each speech feature in all speech features is calculated. The proportion is then proportionally amplified and rounded to obtain the percentage of drunk driving. To increase the weight of the speech, the percentage of drunk driving can be doubled to obtain the initial speech weight.

[0091] In this embodiment, the speech features and the calculated initial speech weights are used as speech feature parameters for constructing the initial drunk driving risk recognition model.

[0092] It should be emphasized that, to further ensure the privacy and security of the feature parameter set, the aforementioned feature parameter set can also be stored in a node of a blockchain.

[0093] The blockchain referred to in this application is a novel application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and encryption algorithms. Essentially, a blockchain is a decentralized database, a chain of data blocks linked together using cryptographic methods. Each data block contains information about a batch of network transactions, used to verify the validity of the information (anti-counterfeiting) and generate the next block. A blockchain can include an underlying blockchain platform, a platform product service layer, and an application service layer.

[0094] It should be noted that the initial speech weights corresponding to the speech features can also be calculated using a pre-trained speech weight model. The training method is the same as above and will not be repeated here.

[0095] Step S204: Combine the basic feature parameters and the speech feature parameters into a feature parameter set, and use the feature parameter set as model parameters to construct an initial drunk driving risk recognition model.

[0096] In this embodiment, an initial drunk driving risk identification model is constructed based on basic feature parameters and voice feature parameters. This initial drunk driving risk identification model can be a mathematical model. A mathematical operation model is constructed based on basic feature parameters and voice feature parameters. For example, the feature parameters are summed (i.e., the weights corresponding to each feature are summed) to obtain the overall score as the score of the drunk driving risk identification model.

[0097] It should be noted that when the same parameter type has multiple feature categories (i.e. multiple keywords, for example, the cause of the accident includes turning and changing lanes, so the keyword with the larger weight value is taken), the keyword with the highest weight is used and is not counted repeatedly. Parameters that do not appear are recorded as having a weight of 0.

[0098] For example: At 23:11:18 on July 15th, a report was received from driver Zhang San (male, 28 years old, the insured in this case). The accident occurred at the bar street of Futian Shopping Park in Shenzhen. The reporter stated that the accident occurred around 22:50, and the cause of the accident was turning. The report also mentioned having dinner with colleagues. The risk score of the drunk driving model is obtained by weighting the above parameters accordingly.

[0099] 5 (South District, July) + 7 (South District, accident occurred between 9 PM and 11 PM) + 10 (Bar Street) + 6 (Driver is male) + 4 (Driver's age is 25-28) + 4 (Time interval between reports is 10-30 minutes) + 2 (The person reporting the accident is the same as the insured) + 5 (Cause of the accident was turning) + 8 (Colleague was mentioned in the accident report) = 45 points.

[0100] In some optional implementations, the drunk driving risk identification model can be a random forest classification model, a Bayesian classification model, or a logistic regression model. It is trained by using basic feature parameters and speech feature parameters to obtain a trained drunk driving risk identification model.

[0101] Step S205: Train the initial drunk driving risk identification model to obtain the target drunk driving risk identification model.

[0102] In this embodiment, drunk driving feature data and drunk driving voice data are combined into a dataset. A training dataset and a validation dataset are extracted from this dataset. The model parameters of the initial drunk driving risk identification model are adjusted based on the training dataset until the model converges to obtain the model to be validated. The validation dataset is input into the model to be validated for validation to obtain the validation result. When the validation result is greater than or equal to a preset threshold, the model to be validated is determined to be the target drunk driving risk identification model.

[0103] Specifically, the training dataset is input into the initial drunk driving risk identification model to obtain drunk driving results. Based on the drunk driving results, the feature contribution of each feature parameter in the model parameters is determined. Based on the feature contribution, the adjustment coefficient of the corresponding feature parameter is determined, and the model parameters are adjusted according to the adjustment coefficient. Iterative training continues until the model is trained to a certain extent. At this point, the model performance reaches the optimal state, and the adjustment coefficient does not change significantly, i.e., convergence.

[0104] In some optional implementations, the steps of determining the feature contribution of each feature parameter in the model parameters based on the drunk driving result, and determining the adjustment coefficient of the corresponding feature parameter based on the feature contribution, include:

[0105] Determine the ratio of each feature parameter to the drunk driving result (i.e., the feature contribution);

[0106] Calculate the difference between the actual score and the drunk driving result, and obtain the adjustment coefficient based on the ratio and the difference.

[0107] For example, suppose the output of the drunk driving result is 5(A) + 4(B) + 3(C) = 12, the actual score is 15, the feature parameter (A, 5) contributes 5 / 12 to the drunk driving result, the feature parameter (B, 4) contributes 4 / 12 to the drunk driving result, the feature parameter (C, 3) contributes 3 / 12 to the drunk driving result, and the difference between the actual score and the drunk driving result is 3. Then the adjustment coefficients are (5 / 12)*3 = 1.25, (4 / 12)*3 = 1 and (3 / 12)*3 = 0.75 respectively. Then the weight of feature A can be adjusted to 5 + 1.25 = 6.25, the weight of feature B to 4 + 1 = 5, and the weight of feature C to 3 + 0.75 = 3.75.

[0108] After the model converges, the model to be verified is output. It is then verified using a verification dataset. If the verification result is greater than or equal to a preset threshold, the model to be verified is output as the target drunk driving risk identification model.

[0109] In this embodiment, by training and validating the drunk driving risk identification model, the final target drunk driving risk identification model is obtained, which can improve the accuracy and efficiency of drunk driving risk identification.

[0110] Step S206: Obtain the target drunk driving characteristics, input the target drunk driving characteristics into the target drunk driving risk identification model, and output the risk prediction result.

[0111] In this embodiment, target drunk driving data is acquired, target drunk driving features are extracted from the target drunk driving data, the target drunk driving features are input into the target drunk driving risk identification model, and the risk prediction result is output.

[0112] In this embodiment, the risk prediction result is a score, which can be set from 0 to 100 points. Different scores correspond to different case labels and drunk driving risk levels, as shown in Table 1.

[0113] Table 1. Relationship between model score and drunk driving risk level

[0114]

[0115] In some optional implementations, user drunk driving assessment information can be generated based on risk prediction results. The corresponding vehicle compensation business can be processed based on the user drunk driving assessment information. After verifying the corresponding vehicle compensation business, the compensation amount corresponding to the vehicle compensation business can be issued to the user. The vehicle compensation business includes vehicle damage compensation business, and may also include personal injury compensation business, which is not limited here.

[0116] This application constructs a drunk driving risk identification model through basic feature parameters and voice feature parameters, realizing risk identification in a structured and multi-factor combined manner. This eliminates the unstable factors caused by the experience of risk control personnel, reduces the false trigger rate, and allows for flexible control of the model's sensitivity by adjusting the parameters during training, thereby improving the efficiency and accuracy of drunk driving risk identification.

[0117] In some optional implementations of this embodiment, the step of obtaining voice features based on drunk driving voice data includes:

[0118] Based on the acoustic features in the drunk driving voice data, the drunk driving voice data is segmented to obtain voice segments;

[0119] Speech recognition processing is performed on the speech segments to obtain the speech recognition results corresponding to the drunk driving speech data;

[0120] Speech features are obtained based on the speech recognition results.

[0121] Specifically, the drunk driving speech data is processed into multiple speech frames, which are then input into a trained speech segmentation model. Based on the acoustic features of the speech frames, the speech frames are segmented to obtain speech segments. The speech segments are then input into a trained speech recognition model, which performs speech recognition processing on the speech segments to convert them into corresponding text. The output text of the speech recognition model is the speech recognition result, from which speech features are extracted.

[0122] The speech segmentation model includes a feature extraction layer, a recognition segmentation layer, and an output layer. The feature extraction layer extracts the acoustic features of the speech frames of the drunk driving speech data; the acoustic features are output to the recognition segmentation layer, and the speech segmentation endpoints are determined based on the segmentation parameters and acoustic features. The speech frame sequence to be recognized is generated based on the speech segmentation endpoints; the speech frame sequence to be recognized is input to the output layer, and the speech segments are output.

[0123] Among them, the identification of the segmentation can be achieved using a bidirectional Bi-LSTM network; the acoustic features include Mel Frequency Cepstrum Coefficient (MFCC), first-order difference MFCC, and second-order difference MFCC, etc.

[0124] Speech recognition of speech data can be achieved using Automatic Speech Recognition (ASR) technology. The speech recognition model can be a pre-acquired traditional speech recognition model or a pre-trained model.

[0125] In this embodiment, the accuracy and efficiency of speech recognition can be improved by using a speech segmentation model and a speech recognition model to perform speech recognition on drunk driving speech data.

[0126] In some optional implementations, the steps described above, which involve inputting the verification dataset into the model to be verified for verification, obtaining the verification result, and determining the model to be verified as the target drunk driving risk identification model when the verification result is greater than or equal to a preset threshold, include:

[0127] Input the validation dataset into the model to be validated to obtain the prediction output;

[0128] Calculate the prediction accuracy based on the prediction output results;

[0129] If the prediction accuracy is greater than or equal to the preset threshold, the model to be verified will be output as the target drunk driving risk identification model.

[0130] If the prediction accuracy is less than a preset threshold, the training dataset is updated, and the step of adjusting the model parameters of the initial drunk driving risk identification model based on the training dataset is executed.

[0131] The process involves inputting the validation dataset into the model to be validated to obtain the predicted output. The prediction accuracy is then calculated based on the predicted output and the actual results. If the prediction accuracy is greater than or equal to a preset threshold, the model to be validated is the target drunk driving risk identification model. If the prediction accuracy is less than the preset threshold, the training dataset is updated, and the initial drunk driving risk identification model is retrained to ensure the model's prediction accuracy.

[0132] This application can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0133] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by instructing related hardware with computer-readable instructions. These computer-readable instructions can be stored in a computer-readable storage medium. When executed, the program can include the processes of the embodiments of the above methods. The aforementioned storage medium can be a non-volatile storage medium such as a magnetic disk, optical disk, or read-only memory (ROM), or random access memory (RAM).

[0134] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0135] Further reference Figure 3 As a response to the above Figure 2 The implementation of the method shown in this application provides an embodiment of a drunk driving risk identification device, which is similar to... Figure 2 Corresponding to the method embodiments shown, this device can be specifically applied to various electronic devices.

[0136] like Figure 3As shown, the drunk driving risk identification device 300 described in this embodiment includes: an acquisition module 301, a basic feature extraction module 302, a voice feature extraction module 303, a construction module 304, a training module 305, and a prediction module 306. Wherein:

[0137] The acquisition module 301 is used to acquire drunk driving cases and to acquire drunk driving feature data and corresponding drunk driving voice data from the drunk driving cases;

[0138] The basic feature extraction module 302 is used to extract basic features from the drunk driving feature data, determine the initial basic weight of the basic features according to the drunk driving case, and obtain basic feature parameters according to the basic features and the corresponding initial basic weights.

[0139] The speech feature extraction module 303 is used to obtain speech features based on the drunk driving speech data, determine the initial speech weight of the speech features, and obtain speech feature parameters based on the speech features and the corresponding initial speech weights.

[0140] The construction module 304 is used to combine the basic feature parameters and the voice feature parameters into a feature parameter set, and use the feature parameter set as model parameters to construct an initial drunk driving risk recognition model;

[0141] Training module 305 is used to train the initial drunk driving risk identification model to obtain the target drunk driving risk identification model;

[0142] The prediction module 306 is used to acquire the target drunk driving characteristics, input the target drunk driving characteristics into the target drunk driving risk identification model, and output the risk prediction result.

[0143] It should be emphasized that, to further ensure the privacy and security of the feature parameter set, the aforementioned feature parameter set can also be stored in a node of a blockchain.

[0144] Based on the aforementioned drunk driving risk identification device, a drunk driving risk identification model is constructed using basic feature parameters and voice feature parameters. This enables risk identification in a structured, multi-factor combined manner, eliminating unstable factors caused by the experience of risk control personnel and reducing the false trigger rate. At the same time, by adjusting the parameters during the training of the drunk driving identification model, the sensitivity of the model can be flexibly controlled, improving the efficiency and accuracy of drunk driving risk identification.

[0145] In this embodiment, the basic feature extraction module 302 includes a statistical submodule, a first calculation submodule, and a second calculation submodule, wherein:

[0146] The statistics submodule is used to count the number of drunk driving cases corresponding to each feature category under the basic features, and obtain the number of categories;

[0147] The first calculation submodule is used to calculate the percentage of drunk driving cases relative to the number of each feature category.

[0148] The second calculation submodule is used to obtain the initial basic weight of the basic feature under each feature category based on the percentage of drunk driving.

[0149] In this embodiment, by calculating the weight of the basic features under each feature category, a structured, multi-factor combined method for identifying drunk driving risks can be achieved, avoiding bias and thus ensuring the stability of the accuracy of drunk driving risk identification.

[0150] In some optional implementations of this embodiment, the speech feature extraction module 303 includes a segmentation submodule, a recognition submodule, and an acquisition submodule, wherein:

[0151] The segmentation module is used to perform speech segmentation processing on the drunk driving speech data based on the acoustic features in the drunk driving speech data to obtain speech segments;

[0152] The recognition submodule is used to perform speech recognition processing on the speech segment to obtain the speech recognition result corresponding to the drunk driving speech data;

[0153] The acquisition submodule is used to obtain speech features based on the speech recognition results.

[0154] This embodiment can improve the accuracy and efficiency of speech recognition.

[0155] In this embodiment, the speech feature extraction module 303 further includes a statistics submodule and a calculation submodule. The statistics submodule is used to count the total number of occurrences of all the speech features and the number of features for each speech feature. The calculation submodule is used to calculate the initial speech weight of the speech feature based on the number of features and the total number of occurrences.

[0156] This embodiment can improve the accuracy and efficiency of determining weights.

[0157] In this embodiment, the training module 305 includes an acquisition submodule, an adjustment submodule, and a verification submodule, wherein:

[0158] The submodule is used to obtain training datasets and validation datasets based on the drunk driving feature data and the drunk driving voice data;

[0159] The adjustment submodule is used to adjust the model parameters of the initial drunk driving risk identification model based on the training dataset until the model converges and the model to be verified is obtained.

[0160] The verification submodule is used to input the verification dataset into the model to be verified for verification, obtain the verification result, and determine the model to be verified as the target drunk driving risk identification model when the verification result is greater than or equal to a preset threshold.

[0161] In this embodiment, by training and validating the drunk driving risk identification model, the final target drunk driving risk identification model is obtained, which can improve the accuracy and efficiency of drunk driving risk identification.

[0162] In this embodiment, the adjustment submodule is further used for:

[0163] The training dataset is input into the initial drunk driving risk identification model to obtain the drunk driving result;

[0164] Determine the feature contribution of each feature parameter in the model parameters based on the drunk driving results;

[0165] The adjustment coefficient of the corresponding feature parameter is determined based on the feature contribution degree;

[0166] The model parameters are adjusted according to the adjustment coefficients.

[0167] In some alternative implementations, the verification submodule is further used for:

[0168] Input the validation dataset into the model to be validated to obtain the prediction output results;

[0169] Calculate the prediction accuracy based on the predicted output results;

[0170] If the prediction accuracy is greater than or equal to a preset threshold, the model to be verified is output as the target drunk driving risk identification model.

[0171] If the prediction accuracy is less than a preset threshold, then the training dataset is updated, and the step of adjusting the model parameters of the initial drunk driving risk identification model based on the training dataset is executed.

[0172] In this embodiment, the accuracy of the model's predictions can be further guaranteed by validating the model.

[0173] To address the aforementioned technical problems, embodiments of this application also provide a computer device. Please refer to [link / reference needed]. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0174] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected via a system bus. It should be noted that only the computer device 4 with components 41-43 is shown in the figure; however, it should be understood that it is not required to implement all the shown components, and more or fewer components can be implemented alternatively. Those skilled in the art will understand that the computer device described here is a device capable of automatically performing numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes, but is not limited to, microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0175] The computer device can be a desktop computer, laptop, handheld computer, or cloud server, etc. The computer device can interact with the user via a keyboard, mouse, remote control, touchpad, or voice control.

[0176] The memory 41 includes at least one type of readable storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 41 may be an internal storage unit of the computer device 4, such as the hard disk or memory of the computer device 4. In other embodiments, the memory 41 may also be an external storage device of the computer device 4, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the computer device 4. Of course, the memory 41 may include both the internal storage unit and its external storage device of the computer device 4. In this embodiment, the memory 41 is typically used to store the operating system and various application software installed on the computer device 4, such as computer-readable instructions for drunk driving risk identification methods. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or will be output.

[0177] In some embodiments, the processor 42 may be a central processing unit (CPU), controller, microcontroller, microprocessor, or other data processing chip. The processor 42 is typically used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to execute computer-readable instructions stored in the memory 41 or to process data, for example, to execute computer-readable instructions for the drunk driving risk identification method.

[0178] The network interface 43 may include a wireless network interface or a wired network interface, which is typically used to establish communication connections between the computer device 4 and other electronic devices.

[0179] This embodiment implements the steps of the drunk driving risk identification method described above by executing computer-readable instructions stored in the memory through the processor. It constructs a drunk driving risk identification model through basic feature parameters and voice feature parameters, and realizes risk identification in a structured and multi-factor combined manner. This eliminates the unstable factors caused by the experience of risk control personnel, reduces the false trigger rate, and at the same time, by adjusting the parameters of the drunk driving identification model during training, the sensitivity of the model can be flexibly controlled, thereby improving the identification efficiency and accuracy of drunk driving risk.

[0180] This application also provides another implementation method, namely, providing a computer-readable storage medium storing computer-readable instructions that can be executed by at least one processor to cause the at least one processor to perform the steps of the drunk driving risk identification method described above. This method constructs a drunk driving risk identification model through basic feature parameters and voice feature parameters, achieving risk identification in a structured, multi-factor combined manner. This eliminates instability caused by the experience of risk control personnel, reducing the false trigger rate. Furthermore, by adjusting the parameters during the training of the drunk driving identification model, the sensitivity of the model can be flexibly controlled, improving the efficiency and accuracy of drunk driving risk identification.

[0181] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0182] Obviously, the embodiments described above are only some embodiments of this application, not all embodiments. The accompanying drawings show preferred embodiments of this application, but do not limit the patent scope of this application. This application can be implemented in many different forms; rather, the purpose of providing these embodiments is to provide a more thorough and comprehensive understanding of the disclosure of this application. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing specific embodiments, or make equivalent substitutions for some of the technical features. Any equivalent structures made using the content of this application's specification and drawings, directly or indirectly applied to other related technical fields, are similarly within the scope of patent protection of this application.

Claims

1. A method of drunk driving risk identification, characterized by, The method comprises the following steps: obtaining a drunk driving case, obtaining drunk driving feature data and corresponding drunk driving voice data from the drunk driving case; extracting basic features from the drunk driving feature data and determining initial basic weights of the basic features according to the drunk driving case, obtaining basic feature parameters according to the basic features and the corresponding initial basic weights; obtaining voice features according to the drunk driving voice data and determining initial voice weights of the voice features, obtaining voice feature parameters according to the voice features and the corresponding initial voice weights; combining the basic feature parameters and the voice feature parameters to form a feature parameter set, taking the feature parameter set as model parameters, and constructing an initial drunk driving risk identification model; training the initial drunk driving risk identification model to obtain a target drunk driving risk identification model; obtaining target drunk driving features, inputting the target drunk driving features into the target drunk driving risk identification model, and outputting a risk prediction result; wherein the step of determining the initial voice weights of the voice features comprises: counting the total number of occurrences of all the voice features and the number of features of each voice feature; and calculating the initial voice weights of the voice features according to the number of features and the total number of occurrences; the step of training the initial drunk driving risk identification model to obtain a target drunk driving risk identification model comprises: obtaining a training data set and a verification data set according to the drunk driving feature data and the drunk driving voice data; adjusting the model parameters of the initial drunk driving risk identification model based on the training data set until the model converges to obtain a to-be-verified model; inputting the verification data set into the to-be-verified model for verification to obtain a verification result, and determining that the to-be-verified model is the target drunk driving risk identification model when the verification result is greater than or equal to a preset threshold; the step of adjusting the model parameters of the initial drunk driving risk identification model based on the training data set comprises: inputting the training data set into the initial drunk driving risk identification model to obtain a drunk driving result; determining the feature contribution degree of each feature parameter in the model parameters according to the drunk driving result, the feature contribution degree being the ratio of the score of the feature parameter to the drunk driving result; obtaining the adjustment coefficient of the corresponding feature parameter based on the difference between the feature contribution degree and the actual score and the drunk driving result; and adjusting the model parameters according to the adjustment coefficient.

2. The drunk driving risk identification method according to claim 1, characterized in that, the step of determining the initial basic weights of the basic features according to the drunk driving case comprises: counting the number of drunk driving cases corresponding to each feature category under the basic features to obtain the number of categories; calculating the drunk driving proportion of the number of categories of each feature category relative to the total number of drunk driving cases; obtaining the initial basic weights of the basic features in each feature category according to the drunk driving proportion.

3. The drunk driving risk identification method of claim 1, wherein, the step of obtaining voice features according to the drunk driving voice data comprises: performing voice segmentation processing on the drunk driving voice data according to the acoustic features in the drunk driving voice data to obtain voice segments; perform voice recognition processing on the voice segment to obtain a voice recognition result corresponding to the drunk driving voice data; obtain voice features according to the voice recognition result.

4. The drunk driving risk identification method of claim 1, wherein, The step of inputting the verification data set into the to-be-verified model for verification to obtain a verification result, and determining that the to-be-verified model is the target drunk driving risk identification model when the verification result is greater than or equal to a preset threshold, comprises: inputting the verification data set into the to-be-verified model to obtain a predicted output result; calculating a prediction accuracy according to the predicted output result; if the prediction accuracy is greater than or equal to a preset threshold, outputting the to-be-verified model as the target drunk driving risk identification model; if the prediction accuracy is less than the preset threshold, updating a training data set and performing the step of adjusting model parameters of the initial drunk driving risk identification model based on the training data set.

5. A drunk driving risk identification apparatus characterized by comprising: comprise: an acquisition module configured to acquire drunk driving cases, and acquire drunk driving feature data and corresponding drunk driving voice data from the drunk driving cases; a basic feature extraction module configured to extract basic features from the drunk driving feature data, determine initial basic weights of the basic features according to the drunk driving cases, and obtain basic feature parameters according to the basic features and the corresponding initial basic weights; a voice feature extraction module configured to acquire voice features according to the drunk driving voice data, determine initial voice weights of the voice features, and obtain voice feature parameters according to the voice features and the corresponding initial voice weights; a construction module configured to combine the basic feature parameters and the voice feature parameters to form a feature parameter set, use the feature parameter set as model parameters, and construct an initial drunk driving risk identification model; a training module configured to train the initial drunk driving risk identification model to obtain a target drunk driving risk identification model; a prediction module configured to acquire target drunk driving features, input the target drunk driving features into the target drunk driving risk identification model, and output a risk prediction result; The voice feature extraction module further comprises a statistical submodule and a calculation submodule. The statistical submodule is configured to count a total number of occurrences of all the voice features and a feature quantity of each voice feature. The calculation submodule is configured to calculate the initial voice weights of the voice features according to the feature quantity and the total number of occurrences. The training module comprises an obtaining submodule, an adjusting submodule, and a verification submodule, wherein: The obtaining submodule is configured to obtain a training data set and a verification data set according to the drunk driving feature data and the drunk driving voice data; The adjusting submodule is configured to adjust model parameters of the initial drunk driving risk identification model based on the training data set until the model converges, to obtain a to-be-verified model; The verification submodule is configured to input the verification data set into the to-be-verified model for verification to obtain a verification result, and determine that the to-be-verified model is the target drunk driving risk identification model when the verification result is greater than or equal to a preset threshold; The adjusting submodule is further configured to: inputting the training data set into the initial drunk driving risk identification model to obtain a drunk driving result; determining a feature contribution degree of each feature parameter in the model parameter according to the drunk driving result, the feature contribution degree being a ratio of a score of the feature parameter to the drunk driving result; obtaining an adjustment coefficient of the corresponding feature parameter based on the feature contribution degree and a difference between an actual score and the drunk driving result; and adjusting the model parameter according to the adjustment coefficient. 6.A computer device, comprising a memory and a processor, wherein the memory stores computer readable instructions, and the processor implements steps of the drunk driving risk identification method according to any one of claims 1 to 4 when executing the computer readable instructions.

7. A computer readable storage medium characterized in that, The computer readable storage medium stores computer readable instructions, and the computer readable instructions implement steps of the drunk driving risk identification method according to any one of claims 1 to 4 when executed by a processor.

Images