An artificial intelligence-based data center abnormal behavior analysis system

Abstract

The application relates to the technical field of an internet car rental platform, in particular to a data center abnormal behavior analysis system based on artificial intelligence, which comprises the following steps: acquiring multi-source car rental data of an internet car rental platform data center; constructing a tenant behavior abnormality identification model to identify tenant behavior data, obtaining a tenant driving abnormality coefficient and a tenant credit abnormality coefficient, and obtaining a tenant behavior abnormality coefficient; constructing a vehicle state abnormality identification model to identify vehicle state data, obtaining a vehicle state abnormality coefficient; constructing a platform operation abnormality identification model to identify platform operation data, obtaining a platform operation abnormality coefficient; obtaining a comprehensive car rental abnormality coefficient according to the three coefficients, obtaining a car rental account abnormality level according to the comprehensive car rental abnormality coefficient, and performing car rental abnormality early warning according to the car rental account abnormality level. The application can effectively improve the accuracy and comprehensiveness of the internet car rental platform data center abnormal behavior analysis, and improve tenant satisfaction and safety.

Description

Technical Field

[0001] This invention relates to the field of internet car rental platform technology, specifically to an artificial intelligence-based data center abnormal behavior analysis system. Background Technology

[0002] With the rapid development of internet technology and the sharing economy, online car rental platforms, as an emerging business model, have gradually become an important mode of transportation in modern cities. Through these platforms, renters can conveniently rent and return vehicles, while the platforms ensure efficient operation through real-time data processing and analysis. However, with the rapid development of these platforms and the increasing demands of renters, online car rental platforms also face many new challenges, especially the analysis of various abnormal behaviors related to renters, vehicles, and the platform itself.

[0003] Currently, traditional internet car rental platform data center anomaly analysis systems mostly focus on single-dimensional data anomaly monitoring and analysis, often failing to fully consider the comprehensive analysis of anomalies involving tenants, vehicles, and the platform. This single-dimensional anomaly analysis method leads to limitations in data center anomaly analysis: it cannot comprehensively analyze the anomalies in the internet car rental platform's data center, resulting in limitations on the efficiency, tenant satisfaction, and security of the internet car rental platform.

[0004] To address this, an artificial intelligence-based data center abnormal behavior analysis system is proposed. Summary of the Invention

[0005] The purpose of this invention is to provide an AI-based data center abnormal behavior analysis system. This system acquires multi-source car rental data from an internet car rental platform's data center; constructs a tenant behavior anomaly identification model to identify tenant behavior data, obtaining tenant driving anomaly coefficients and tenant credit anomaly coefficients; constructs a vehicle status anomaly identification model to identify vehicle status data, obtaining vehicle status anomaly coefficients; and constructs a platform operation anomaly identification model to identify platform operation data, obtaining platform operation anomaly coefficients. Based on the tenant behavior anomaly coefficient, vehicle status anomaly coefficient, and platform operation anomaly coefficient, a comprehensive car rental anomaly coefficient is obtained. Based on the comprehensive car rental anomaly coefficient, a car rental account anomaly level is determined, and a car rental anomaly warning is issued based on the car rental account anomaly level. This invention can effectively improve the accuracy and comprehensiveness of abnormal behavior analysis in internet car rental platform data centers, enhancing tenant satisfaction and security.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] An artificial intelligence-based data center abnormal behavior analysis system includes:

[0008] The car rental data acquisition module is used to acquire multi-source car rental data from the data center of the internet car rental platform in real time. The multi-source car rental data includes the tenant's historical behavior data, historical vehicle status data, and the tenant's historical platform operation data for rented vehicles under a single car rental account.

[0009] The tenant abnormal behavior identification module is used to build a tenant behavior abnormality identification model to identify tenant behavior data, obtain the tenant driving abnormality coefficient and the tenant credit abnormality coefficient, and then obtain the tenant behavior abnormality coefficient.

[0010] The vehicle anomaly identification module is used to build a vehicle state anomaly identification model to identify vehicle state data and obtain vehicle state anomaly coefficients.

[0011] The platform operation anomaly identification module is used to build a platform operation anomaly identification model to identify platform operation data and obtain platform operation anomaly coefficients;

[0012] The comprehensive anomaly analysis module is used to obtain a comprehensive car rental anomaly coefficient based on the tenant behavior anomaly coefficient, vehicle status anomaly coefficient and platform operation anomaly coefficient, obtain a car rental account anomaly level based on the comprehensive car rental anomaly coefficient and a first preset threshold set, and issue a car rental anomaly warning based on the car rental account anomaly level.

[0013] Preferably, the tenant behavior data includes tenant driving data and tenant credit data; the tenant driving data includes driving speed logs and driving process images; the driving process images include driving road condition images and tenant driving images; the tenant credit data includes vehicle return process logs and payment process logs; the vehicle status data includes fault reporting logs, vehicle location logs, and vehicle occupancy logs; and the platform operation data includes fault handling logs and car rental order processing logs.

[0014] Preferably, the tenant behavior anomaly identification model includes a tenant behavior data input layer, a tenant behavior data preprocessing layer, a tenant anomaly feature extraction layer, a tenant behavior anomaly analysis layer, and a tenant behavior anomaly coefficient output layer;

[0015] The tenant behavior data input layer is used to input tenant driving data and tenant credit data into the tenant behavior anomaly identification model; the tenant behavior data preprocessing layer is used to preprocess the tenant driving data and tenant credit data, including data cleaning of driving start time, driving end time, driving speed, vehicle return process logs, and payment process logs; and grayscale conversion and noise reduction of driving process images; the tenant behavior anomaly feature extraction layer is used to extract features from the preprocessed tenant driving data and tenant credit data, including detecting driving road condition images and tenant driving images using the YOLOv8 object detection algorithm, and using LSTM to detect driving speed daily. The system extracts features from the tenant's credit data to obtain tenant driving anomaly features, including the frequency of illegal lane changes, red light running, fatigue driving, and speeding. It then extracts features from the preprocessed tenant credit data using LSTM to obtain tenant credit anomaly features, including the frequency of delayed vehicle return, delayed payment, and payment failure. The tenant behavior anomaly analysis layer obtains a tenant driving anomaly coefficient based on the tenant driving anomaly features. Finally, it obtains a tenant behavior anomaly coefficient based on the tenant credit anomaly features. The tenant behavior anomaly coefficient output layer outputs the tenant behavior anomaly coefficient.

[0016] Preferably, the tenant behavior anomaly coefficient is:

[0017] Using the tenant behavior anomaly impact coefficient as a weight, the weighted sum of the tenant driving anomaly coefficient and the tenant credit anomaly coefficient is first calculated and used as the input of the exponential function to obtain the aforementioned tenant behavior anomaly coefficient.

[0018] Preferably, the vehicle status anomaly recognition model includes a vehicle status data input layer, a vehicle status data preprocessing layer, a vehicle status anomaly feature extraction layer, a vehicle status anomaly feature analysis layer, and a vehicle status anomaly coefficient output layer. The vehicle status data input layer is used to input the vehicle status data into the vehicle status anomaly recognition model. The vehicle status data preprocessing layer is used to clean the vehicle status data. The vehicle status anomaly feature extraction layer is used to analyze the preprocessed vehicle status data using a CNN to obtain vehicle status anomaly features. The vehicle status anomaly features include false alarm frequency, vehicle location anomaly frequency, and vehicle abnormal occupancy frequency. The vehicle status anomaly feature analysis layer is used to obtain vehicle status anomaly coefficients based on the vehicle status anomaly features. The vehicle status anomaly coefficient output layer is used to output the vehicle status anomaly coefficients.

[0019] Preferably, the platform operation anomaly identification model includes a platform operation data input layer, a platform operation data preprocessing layer, a platform operation anomaly feature extraction layer, a platform operation anomaly feature analysis layer, and a platform operation anomaly coefficient output layer;

[0020] The platform operation data input layer is used to input the platform operation data into the platform operation anomaly identification model; the platform operation data preprocessing layer is used to preprocess the platform operation data; the platform operation anomaly feature extraction layer is used to extract features from the preprocessed platform operation data using CNN to obtain platform operation anomaly features; the platform operation anomaly features include average car rental fault response time, average car rental order processing time, and average car rental order cancellation processing time; the platform operation anomaly feature analysis layer is used to obtain platform operation anomaly coefficients based on the platform operation anomaly features; and the platform operation anomaly coefficient output layer is used to output the platform operation anomaly coefficients.

[0021] The preferred overall car rental anomaly coefficient is:

[0022] The comprehensive car rental anomaly coefficient is obtained by weighting the tenant behavior anomaly coefficient, vehicle status anomaly coefficient, and platform operation anomaly coefficient by using the car rental anomaly impact coefficient as a weight.

[0023] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0024] 1. This invention acquires multi-dimensional car rental data, including historical tenant behavior data, historical vehicle status data, and historical platform operation data of tenants renting vehicles. By conducting comprehensive multi-dimensional analysis of tenants, vehicles, and the platform based on the internet car rental platform, it can effectively improve the efficiency of the internet car rental platform and enhance tenant satisfaction and security.

[0025] 2. This invention identifies tenant behavior data by constructing a tenant behavior anomaly identification model, obtains tenant driving anomaly coefficient and tenant credit anomaly coefficient, and then obtains tenant behavior anomaly coefficient; through comprehensive analysis based on tenant behavior anomaly and tenant credit anomaly, the tenant behavior anomaly coefficient is obtained. The tenant behavior anomaly identification model can help improve the comprehensiveness of abnormal behavior analysis of the data center of the Internet car rental platform, thereby effectively improving the efficiency of the Internet car rental platform and improving tenant satisfaction and security.

[0026] 3. This invention identifies vehicle status data by constructing a vehicle status anomaly identification model to obtain vehicle status anomaly features; and obtains vehicle status anomaly coefficients based on the vehicle status anomaly features; based on the vehicle status anomaly coefficients, it can help improve the comprehensiveness of abnormal behavior analysis of the data center of the Internet car rental platform, thereby effectively improving the efficiency of the Internet car rental platform and enhancing tenant satisfaction and security.

[0027] 4. This invention analyzes platform operation data by constructing a platform operation anomaly identification model package; it extracts features from the preprocessed platform operation data using CNN to obtain platform operation anomaly features; these features include average car rental fault response time, average car rental order processing time, and average car rental order cancellation processing time; and it obtains the platform operation anomaly coefficient based on these features. By constructing the platform operation anomaly identification model, the comprehensiveness of abnormal behavior analysis of the internet car rental platform data center can be improved, thereby effectively improving the efficiency of the internet car rental platform and enhancing tenant satisfaction and security.

[0028] 5. This invention obtains a comprehensive car rental anomaly coefficient based on tenant behavior anomaly coefficient, vehicle status anomaly coefficient, and platform operation anomaly coefficient. Based on the comprehensive car rental anomaly coefficient, a car rental account anomaly level is obtained. Car rental anomaly warnings are issued based on the car rental account anomaly level. This helps to improve the comprehensiveness of abnormal behavior analysis in the data center of the Internet car rental platform, thereby effectively improving the efficiency of the Internet car rental platform and enhancing tenant satisfaction and security. Attached Figure Description

[0029] Figure 1 A schematic diagram of the structure of an artificial intelligence-based data center abnormal behavior analysis system provided in an embodiment of the present invention;

[0030] Figure 2 This is a schematic diagram of the structure of a tenant behavior anomaly identification model provided in an embodiment of the present invention;

[0031] Figure 3 This is a schematic diagram of the structure of a vehicle state anomaly recognition model provided in an embodiment of the present invention;

[0032] Figure 4 This is a schematic diagram of the structure of a platform operation anomaly identification model provided in an embodiment of the present invention. Detailed Implementation

[0033] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0034] Example 1

[0035] To improve the platform efficiency, tenant satisfaction, and security of Internet car rental platform A, an AI-based data center abnormal behavior analysis system was applied.

[0036] Reference Figure 1 The diagram below illustrates the structure of an artificial intelligence-based data center abnormal behavior analysis system according to an embodiment of the present invention, comprising:

[0037] The car rental data acquisition module is used to acquire multi-source car rental data from the data center of the internet car rental platform in real time. The multi-source car rental data includes the tenant's historical behavior data, historical vehicle status data, and the tenant's historical platform operation data for rented vehicles under a single car rental account.

[0038] Furthermore, the tenant behavior data includes tenant driving data and tenant credit data; the tenant driving data includes driving speed logs and driving process images; the driving process images include driving road condition images and tenant driving images; the tenant credit data includes return process logs and payment process logs; the vehicle status data includes fault reporting logs, vehicle location logs, and vehicle occupancy logs; the platform operation data includes fault handling logs and car rental order processing logs.

[0039] This embodiment acquires multi-dimensional car rental data, including tenant historical behavior data, historical vehicle status data, and historical platform operation data of tenants renting vehicles. By conducting comprehensive multi-dimensional analysis of tenants, vehicles, and the platform based on the internet car rental platform, the efficiency of the internet car rental platform can be effectively improved, and tenant satisfaction and safety can be enhanced.

[0040] The tenant abnormal behavior identification module is used to build a tenant behavior abnormality identification model to identify tenant behavior data, obtain the tenant driving abnormality coefficient and the tenant credit abnormality coefficient, and then obtain the tenant behavior abnormality coefficient.

[0041] Furthermore, the tenant behavior anomaly identification model includes a tenant behavior data input layer, a tenant behavior data preprocessing layer, a tenant anomaly feature extraction layer, a tenant behavior anomaly analysis layer, and a tenant behavior anomaly coefficient output layer; such as Figure 2 This is a schematic diagram of the structure of a tenant behavior anomaly identification model provided in an embodiment of the present invention;

[0042] The tenant behavior data input layer is used to input tenant driving data and tenant credit data into the tenant behavior anomaly identification model; the tenant behavior data preprocessing layer is used to preprocess the tenant driving data and tenant credit data, including data cleaning of driving start time, driving end time, driving speed, vehicle return process logs, and payment process logs; and grayscale conversion and noise reduction of driving process images; the tenant behavior anomaly feature extraction layer is used to extract features from the preprocessed tenant driving data and tenant credit data, including detecting driving road condition images and tenant driving images using the YOLOv8 object detection algorithm, and using LSTM to detect driving speed daily. The system extracts features from the tenant's credit data to obtain tenant driving anomaly features, including the frequency of illegal lane changes, red light running, fatigue driving, and speeding. It then extracts features from the preprocessed tenant credit data using LSTM to obtain tenant credit anomaly features, including the frequency of delayed vehicle return, delayed payment, and payment failure. The tenant behavior anomaly analysis layer obtains a tenant driving anomaly coefficient based on the tenant driving anomaly features. Finally, it obtains a tenant behavior anomaly coefficient based on the tenant credit anomaly features. The tenant behavior anomaly coefficient output layer outputs the tenant behavior anomaly coefficient.

[0043] Furthermore, the tenant driving anomaly coefficient is:

[0044] Using the tenant's abnormal driving impact coefficient as a weight, the frequency of illegal lane changes, red light running, fatigue driving, and speeding driving are weighted and summed, and the square root is taken to obtain the tenant's abnormal driving coefficient.

[0045] Furthermore, the tenant credit anomaly coefficient is:

[0046] The tenant credit anomaly impact coefficient is obtained by using the tenant credit anomaly impact coefficient as a weight to weight the frequency of delayed vehicle return, the frequency of delayed payment, and the frequency of payment failure, and then taking the square root.

[0047] Furthermore, the tenant behavior anomaly coefficient is:

[0048] Using the tenant behavior anomaly impact coefficient as a weight, the weighted sum of the tenant driving anomaly coefficient and the tenant credit anomaly coefficient is first calculated and used as the input of the exponential function to obtain the aforementioned tenant behavior anomaly coefficient.

[0049] This embodiment identifies tenant behavior data by constructing a tenant behavior anomaly identification model, obtaining tenant driving anomaly coefficient and tenant credit anomaly coefficient, and then obtaining tenant behavior anomaly coefficient; through comprehensive analysis based on tenant behavior anomaly and tenant credit anomaly, the tenant behavior anomaly coefficient is obtained. The tenant behavior anomaly identification model can help improve the comprehensiveness of abnormal behavior analysis in the data center of the Internet car rental platform, thereby effectively improving the efficiency of the Internet car rental platform and enhancing tenant satisfaction and security.

[0050] The vehicle anomaly detection module is used to build a vehicle state anomaly detection model to identify vehicle state data and obtain vehicle state anomaly coefficients; for example... Figure 3 This is a schematic diagram of the structure of a vehicle state anomaly recognition model provided in an embodiment of the present invention;

[0051] Furthermore, the vehicle status anomaly recognition model includes a vehicle status data input layer, a vehicle status data preprocessing layer, a vehicle status anomaly feature extraction layer, a vehicle status anomaly feature analysis layer, and a vehicle status anomaly coefficient output layer. The vehicle status data input layer is used to input the vehicle status data into the vehicle status anomaly recognition model. The vehicle status data preprocessing layer is used to clean the vehicle status data. The vehicle status anomaly feature extraction layer is used to analyze the preprocessed vehicle status data using a CNN to obtain vehicle status anomaly features. The vehicle status anomaly features include false alarm frequency, vehicle location anomaly frequency, and vehicle abnormal occupancy frequency. The vehicle status anomaly feature analysis layer is used to obtain vehicle status anomaly coefficients based on the vehicle status anomaly features. The vehicle status anomaly coefficient output layer is used to output the vehicle status anomaly coefficients.

[0052] Furthermore, the vehicle state anomaly coefficient is:

[0053] Using the vehicle status anomaly impact coefficient as a weight, the frequency of false alarms, the frequency of abnormal vehicle location, and the frequency of abnormal vehicle occupancy are weighted and summed, and the square root is taken to obtain the vehicle status anomaly coefficient.

[0054] This embodiment identifies vehicle status data by constructing a vehicle status anomaly recognition model to obtain vehicle status anomaly features; based on the vehicle status anomaly features, a vehicle status anomaly coefficient is obtained; based on the vehicle status anomaly coefficient, it can help improve the comprehensiveness of abnormal behavior analysis of the data center of the Internet car rental platform, thereby effectively improving the efficiency of the Internet car rental platform and enhancing tenant satisfaction and security.

[0055] The platform operation anomaly identification module is used to build a platform operation anomaly identification model to identify platform operation data and obtain platform operation anomaly coefficients; for example... Figure 4This is a schematic diagram of the structure of a platform operation anomaly identification model provided in an embodiment of the present invention;

[0056] Furthermore, the platform operation anomaly identification model includes a platform operation data input layer, a platform operation data preprocessing layer, a platform operation anomaly feature extraction layer, a platform operation anomaly feature analysis layer, and a platform operation anomaly coefficient output layer;

[0057] The platform operation data input layer is used to input the platform operation data into the platform operation anomaly identification model; the platform operation data preprocessing layer is used to preprocess the platform operation data; the platform operation anomaly feature extraction layer is used to extract features from the preprocessed platform operation data using CNN to obtain platform operation anomaly features; the platform operation anomaly features include average car rental fault response time, average car rental order processing time, and average car rental order cancellation processing time; the platform operation anomaly feature analysis layer is used to obtain platform operation anomaly coefficients based on the platform operation anomaly features; and the platform operation anomaly coefficient output layer is used to output the platform operation anomaly coefficients.

[0058] Furthermore, the platform's operational anomaly coefficient is:

[0059] Using the vehicle status anomaly impact coefficient as a weight, the average car rental fault response time, average car rental order processing time, and average car rental order cancellation processing time are weighted and summed, and the square root is used to obtain the platform operation anomaly coefficient.

[0060] This embodiment analyzes platform operation data by constructing a platform operation anomaly identification model package; it extracts features from the preprocessed platform operation data using CNN to obtain platform operation anomaly features; these features include average car rental fault response time, average car rental order processing time, and average car rental order cancellation processing time; and it obtains the platform operation anomaly coefficient based on these features. By constructing the platform operation anomaly identification model, the comprehensiveness of abnormal behavior analysis of the internet car rental platform's data center can be improved, thereby effectively improving the efficiency of the internet car rental platform and enhancing tenant satisfaction and security.

[0061] The comprehensive anomaly analysis module is used to obtain a comprehensive car rental anomaly coefficient based on the tenant behavior anomaly coefficient, vehicle status anomaly coefficient and platform operation anomaly coefficient, obtain a car rental account anomaly level based on the comprehensive car rental anomaly coefficient and a first preset threshold set, and issue a car rental anomaly warning based on the car rental account anomaly level.

[0062] Furthermore, the overall car rental anomaly coefficient is:

[0063] The comprehensive car rental anomaly coefficient is obtained by weighting the tenant behavior anomaly coefficient, vehicle status anomaly coefficient, and platform operation anomaly coefficient by using the car rental anomaly impact coefficient as a weight.

[0064] This embodiment obtains a comprehensive car rental anomaly coefficient based on tenant behavior anomaly coefficient, vehicle status anomaly coefficient, and platform operation anomaly coefficient. Based on the comprehensive car rental anomaly coefficient, the car rental account anomaly level is obtained. Anomaly warnings are issued based on the car rental account anomaly level, which helps to improve the comprehensiveness of abnormal behavior analysis in the data center of the internet car rental platform. This can effectively improve the efficiency of the internet car rental platform and enhance tenant satisfaction and security.

[0065] Furthermore, the abnormality levels of car rental accounts are categorized into severely abnormal, moderately abnormal, slightly abnormal, and normal accounts; warnings are issued for severely abnormal, moderately abnormal, and slightly abnormal accounts.

[0066] To verify the effectiveness of the AI-based data center anomaly behavior analysis system proposed in this embodiment in improving the platform efficiency, tenant satisfaction, and security of Internet car rental platform A, the frequency of accidents and changes in tenant satisfaction were analyzed by comparing the application of different systems on Internet car rental platform A with the original Internet car rental platform. The specific results are shown in Table 1. System 1 is the AI-based data center anomaly behavior analysis system provided in this embodiment; System 2 is a comprehensive anomaly analysis based on System 1 without considering tenants, vehicles, and the platform; System 3 is a comprehensive anomaly analysis based on System 1 without considering anomaly warnings.

[0067] Table 1. Accident frequency and tenant satisfaction of Internet car rental platform A under different system applications.

[0068] system Frequency of accidents Tenant satisfaction System 1 Down 12% Upgrade 18 System 2 down 9% Upgrade 12 System 3 7% decrease Upgrade 9

[0069] As shown in Table 1, the data center abnormal behavior analysis system based on artificial intelligence provided in this embodiment has a certain degree of effectiveness.

[0070] This embodiment acquires multi-source car rental data from the data center of an online car rental platform; constructs a tenant behavior anomaly identification model to identify tenant behavior data, obtaining tenant driving anomaly coefficients and tenant credit anomaly coefficients, and thus a tenant behavior anomaly coefficient; constructs a vehicle status anomaly identification model to identify vehicle status data, obtaining a vehicle status anomaly coefficient; constructs a platform operation anomaly identification model to identify platform operation data, obtaining a platform operation anomaly coefficient; based on these three coefficients, a comprehensive car rental anomaly coefficient is obtained; based on the comprehensive car rental anomaly coefficient, a car rental account anomaly level is obtained; and based on the car rental account anomaly level, a car rental anomaly warning is issued. This embodiment can effectively improve the accuracy and comprehensiveness of anomaly behavior analysis in the data center of an online car rental platform, improving tenant satisfaction and security.

[0071] Example 2

[0072] To improve the platform efficiency, tenant satisfaction, and security of the B2B car rental platform, an AI-based data center abnormal behavior analysis system was applied.

[0073] Reference Figure 1 The diagram below illustrates the structure of an artificial intelligence-based data center abnormal behavior analysis system according to an embodiment of the present invention, comprising:

[0074] The car rental data acquisition module is used to acquire multi-source car rental data from the data center of the internet car rental platform in real time. The multi-source car rental data includes the tenant's historical behavior data, historical vehicle status data, and the tenant's historical platform operation data for rented vehicles under a single car rental account.

[0075] Furthermore, the tenant behavior data includes tenant driving data and tenant credit data; the tenant driving data includes driving speed logs and driving process images; the driving process images include driving road condition images and tenant driving images; the tenant credit data includes return process logs and payment process logs; the vehicle status data includes fault reporting logs, vehicle location logs, and vehicle occupancy logs; the platform operation data includes fault handling logs and car rental order processing logs.

[0076] This embodiment acquires multi-dimensional car rental data, including tenant historical behavior data, historical vehicle status data, and historical platform operation data of tenants renting vehicles. By conducting comprehensive multi-dimensional analysis of tenants, vehicles, and the platform based on the internet car rental platform, the efficiency of the internet car rental platform can be effectively improved, and tenant satisfaction and safety can be enhanced.

[0077] The tenant abnormal behavior identification module is used to build a tenant behavior abnormality identification model to identify tenant behavior data, obtain the tenant driving abnormality coefficient and the tenant credit abnormality coefficient, and then obtain the tenant behavior abnormality coefficient.

[0078] Furthermore, the tenant behavior anomaly identification model includes a tenant behavior data input layer, a tenant behavior data preprocessing layer, a tenant anomaly feature extraction layer, a tenant behavior anomaly analysis layer, and a tenant behavior anomaly coefficient output layer; such as Figure 2 This is a schematic diagram of the structure of a tenant behavior anomaly identification model provided in an embodiment of the present invention;

[0079] The tenant behavior data input layer is used to input tenant driving data and tenant credit data into the tenant behavior anomaly identification model; the tenant behavior data preprocessing layer is used to preprocess the tenant driving data and tenant credit data, including data cleaning of driving start time, driving end time, driving speed, vehicle return process logs, and payment process logs; and grayscale conversion and noise reduction of driving process images; the tenant behavior anomaly feature extraction layer is used to extract features from the preprocessed tenant driving data and tenant credit data, including detecting driving road condition images and tenant driving images using the YOLOv8 object detection algorithm, and using LSTM to detect driving speed daily. The system extracts features from the tenant's credit data to obtain tenant driving anomaly features, including the frequency of illegal lane changes, red light running, fatigue driving, and speeding. It then extracts features from the preprocessed tenant credit data using LSTM to obtain tenant credit anomaly features, including the frequency of delayed vehicle return, delayed payment, and payment failure. The tenant behavior anomaly analysis layer obtains a tenant driving anomaly coefficient based on the tenant driving anomaly features. Finally, it obtains a tenant behavior anomaly coefficient based on the tenant credit anomaly features. The tenant behavior anomaly coefficient output layer outputs the tenant behavior anomaly coefficient.

[0080] Furthermore, the tenant driving anomaly coefficient is:

[0081] Using the tenant's abnormal driving impact coefficient as a weight, the frequency of illegal lane changes, red light running, fatigue driving, and speeding driving are weighted and summed, and the square root is taken to obtain the tenant's abnormal driving coefficient.

[0082] Furthermore, the tenant credit anomaly coefficient is:

[0083] The tenant credit anomaly impact coefficient is obtained by using the tenant credit anomaly impact coefficient as a weight to weight the frequency of delayed vehicle return, the frequency of delayed payment, and the frequency of payment failure, and then taking the square root.

[0084] Furthermore, the tenant behavior anomaly coefficient is:

[0085] Using the tenant behavior anomaly impact coefficient as a weight, the weighted sum of the tenant driving anomaly coefficient and the tenant credit anomaly coefficient is first calculated and used as the input of the exponential function to obtain the aforementioned tenant behavior anomaly coefficient.

[0086] This embodiment identifies tenant behavior data by constructing a tenant behavior anomaly identification model, obtaining tenant driving anomaly coefficient and tenant credit anomaly coefficient, and then obtaining tenant behavior anomaly coefficient; through comprehensive analysis based on tenant behavior anomaly and tenant credit anomaly, the tenant behavior anomaly coefficient is obtained. The tenant behavior anomaly identification model can help improve the comprehensiveness of abnormal behavior analysis in the data center of the Internet car rental platform, thereby effectively improving the efficiency of the Internet car rental platform and enhancing tenant satisfaction and security.

[0087] The vehicle anomaly detection module is used to build a vehicle state anomaly detection model to identify vehicle state data and obtain vehicle state anomaly coefficients; for example... Figure 3 This is a schematic diagram of the structure of a vehicle state anomaly recognition model provided in an embodiment of the present invention;

[0088] Furthermore, the vehicle status anomaly recognition model includes a vehicle status data input layer, a vehicle status data preprocessing layer, a vehicle status anomaly feature extraction layer, a vehicle status anomaly feature analysis layer, and a vehicle status anomaly coefficient output layer. The vehicle status data input layer is used to input the vehicle status data into the vehicle status anomaly recognition model. The vehicle status data preprocessing layer is used to clean the vehicle status data. The vehicle status anomaly feature extraction layer is used to analyze the preprocessed vehicle status data using a CNN to obtain vehicle status anomaly features. The vehicle status anomaly features include false alarm frequency, vehicle location anomaly frequency, and vehicle abnormal occupancy frequency. The vehicle status anomaly feature analysis layer is used to obtain vehicle status anomaly coefficients based on the vehicle status anomaly features. The vehicle status anomaly coefficient output layer is used to output the vehicle status anomaly coefficients.

[0089] Furthermore, the vehicle state anomaly coefficient is:

[0090] Using the vehicle status anomaly impact coefficient as a weight, the frequency of false alarms, the frequency of abnormal vehicle location, and the frequency of abnormal vehicle occupancy are weighted and summed, and the square root is taken to obtain the vehicle status anomaly coefficient.

[0091] This embodiment identifies vehicle status data by constructing a vehicle status anomaly recognition model to obtain vehicle status anomaly features; based on the vehicle status anomaly features, a vehicle status anomaly coefficient is obtained; based on the vehicle status anomaly coefficient, it can help improve the comprehensiveness of abnormal behavior analysis of the data center of the Internet car rental platform, thereby effectively improving the efficiency of the Internet car rental platform and enhancing tenant satisfaction and security.

[0092] The platform operation anomaly identification module is used to build a platform operation anomaly identification model to identify platform operation data and obtain platform operation anomaly coefficients; for example... Figure 4This is a schematic diagram of the structure of a platform operation anomaly identification model provided in an embodiment of the present invention;

[0093] Furthermore, the platform operation anomaly identification model includes a platform operation data input layer, a platform operation data preprocessing layer, a platform operation anomaly feature extraction layer, a platform operation anomaly feature analysis layer, and a platform operation anomaly coefficient output layer;

[0094] The platform operation data input layer is used to input the platform operation data into the platform operation anomaly identification model; the platform operation data preprocessing layer is used to preprocess the platform operation data; the platform operation anomaly feature extraction layer is used to extract features from the preprocessed platform operation data using CNN to obtain platform operation anomaly features; the platform operation anomaly features include average car rental fault response time, average car rental order processing time, and average car rental order cancellation processing time; the platform operation anomaly feature analysis layer is used to obtain platform operation anomaly coefficients based on the platform operation anomaly features; and the platform operation anomaly coefficient output layer is used to output the platform operation anomaly coefficients.

[0095] Furthermore, the platform's operational anomaly coefficient is:

[0096] Using the vehicle status anomaly impact coefficient as a weight, the average car rental fault response time, average car rental order processing time, and average car rental order cancellation processing time are weighted and summed, and the square root is used to obtain the platform operation anomaly coefficient.

[0097] This embodiment analyzes platform operation data by constructing a platform operation anomaly identification model package; it extracts features from the preprocessed platform operation data using CNN to obtain platform operation anomaly features; these features include average car rental fault response time, average car rental order processing time, and average car rental order cancellation processing time; and it obtains the platform operation anomaly coefficient based on these features. By constructing the platform operation anomaly identification model, the comprehensiveness of abnormal behavior analysis of the internet car rental platform's data center can be improved, thereby effectively improving the efficiency of the internet car rental platform and enhancing tenant satisfaction and security.

[0098] The comprehensive anomaly analysis module is used to obtain a comprehensive car rental anomaly coefficient based on the tenant behavior anomaly coefficient, vehicle status anomaly coefficient and platform operation anomaly coefficient, obtain a car rental account anomaly level based on the comprehensive car rental anomaly coefficient and a first preset threshold set, and issue a car rental anomaly warning based on the car rental account anomaly level.

[0099] Furthermore, the overall car rental anomaly coefficient is:

[0100] The comprehensive car rental anomaly coefficient is obtained by weighting the tenant behavior anomaly coefficient, vehicle status anomaly coefficient, and platform operation anomaly coefficient by using the car rental anomaly impact coefficient as a weight.

[0101] This embodiment obtains a comprehensive car rental anomaly coefficient based on tenant behavior anomaly coefficient, vehicle status anomaly coefficient, and platform operation anomaly coefficient. Based on the comprehensive car rental anomaly coefficient, the car rental account anomaly level is obtained. Anomaly warnings are issued based on the car rental account anomaly level, which helps to improve the comprehensiveness of abnormal behavior analysis in the data center of the internet car rental platform. This can effectively improve the efficiency of the internet car rental platform and enhance tenant satisfaction and security.

[0102] Furthermore, the abnormality levels of car rental accounts are categorized into severely abnormal, moderately abnormal, slightly abnormal, and normal accounts; warnings are issued for severely abnormal, moderately abnormal, and slightly abnormal accounts.

[0103] To verify the effectiveness of the AI-based data center anomaly behavior analysis system proposed in this embodiment in improving the platform efficiency, tenant satisfaction, and security of a B-end car rental platform, the frequency of accidents and changes in tenant satisfaction were analyzed by comparing the B-end car rental platform with different systems and the original B-end car rental platform. The specific results are shown in Table 2. System 1 is the AI-based data center anomaly behavior analysis system provided in this embodiment; System 2 is a comprehensive anomaly analysis based on System 1 without considering tenants, vehicles, and the platform; System 3 is a comprehensive anomaly analysis based on System 1 without considering anomaly warnings.

[0104] Table 2. Accident frequency and tenant satisfaction of B2B car rental platforms under different system applications.

[0105] system Frequency of accidents Tenant satisfaction System 1 Down 14% Increase by 20 System 2 down 9% Upgrade 11 System 3 down 8% Improve 10

[0106] As shown in Table 2, the data center abnormal behavior analysis system based on artificial intelligence provided in this embodiment has a certain degree of effectiveness.

[0107] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An artificial intelligence-based data center abnormal behavior analysis system, characterized in that, include: The car rental data acquisition module is used to acquire multi-source car rental data from the data center of the internet car rental platform in real time. The multi-source car rental data includes the tenant's historical behavior data, historical vehicle status data, and the tenant's historical platform operation data for rented vehicles under a single car rental account. The tenant abnormal behavior identification module is used to build a tenant behavior abnormality identification model to identify tenant behavior data, obtain the tenant driving abnormality coefficient and the tenant credit abnormality coefficient, and then obtain the tenant behavior abnormality coefficient. The vehicle anomaly identification module is used to build a vehicle state anomaly identification model to identify vehicle state data and obtain vehicle state anomaly coefficients. The platform operation anomaly identification module is used to construct a platform operation anomaly identification model to identify platform operation data and obtain platform operation anomaly coefficients. The platform operation anomaly identification model includes a platform operation data input layer, a platform operation data preprocessing layer, a platform operation anomaly feature extraction layer, a platform operation anomaly feature analysis layer, and a platform operation anomaly coefficient output layer. The platform operation data input layer is used to input the platform operation data into the platform operation anomaly identification model; the platform operation data preprocessing layer is used to preprocess the platform operation data; the platform operation anomaly feature extraction layer is used to extract features from the preprocessed platform operation data using CNN to obtain platform operation anomaly features; the platform operation anomaly features include average car rental fault response time, average car rental order processing time, and average car rental order cancellation processing time; the platform operation anomaly feature analysis layer is used to obtain the platform operation anomaly coefficient based on the platform operation anomaly features; the platform operation anomaly coefficient output layer is used to output the platform operation anomaly coefficient; The comprehensive anomaly analysis module is used to obtain a comprehensive car rental anomaly coefficient based on the tenant behavior anomaly coefficient, vehicle status anomaly coefficient and platform operation anomaly coefficient, obtain a car rental account anomaly level based on the comprehensive car rental anomaly coefficient and a first preset threshold set, and issue a car rental anomaly warning based on the car rental account anomaly level.

2. The data center abnormal behavior analysis system based on artificial intelligence according to claim 1, characterized in that: Tenant behavior data includes tenant driving data and tenant credit data; tenant driving data includes driving speed logs and driving process images; driving process images include driving road condition images and tenant driving images; Tenant credit data includes vehicle return process logs and payment process logs; vehicle status data includes fault reporting logs, vehicle location logs, and vehicle occupancy logs; platform operation data includes fault handling logs and car rental order processing logs.

3. The data center abnormal behavior analysis system based on artificial intelligence according to claim 2, characterized in that: The tenant behavior anomaly identification model includes a tenant behavior data input layer, a tenant behavior data preprocessing layer, a tenant anomaly feature extraction layer, a tenant behavior anomaly analysis layer, and a tenant behavior anomaly coefficient output layer. The tenant behavior data input layer is used to input tenant driving data and tenant credit data into the tenant behavior anomaly identification model; the tenant behavior data preprocessing layer is used to preprocess tenant driving data and tenant credit data, including data cleaning of driving start time, driving end time, driving speed, vehicle return process log and payment process log; The driving process images are grayscaled and denoised; the tenant behavior abnormal feature extraction layer is used to extract features from the preprocessed tenant driving data and tenant credit data, including detecting driving road condition images and tenant driving images through the YOLOv8 target detection algorithm, and extracting features from driving speed logs through LSTM, thereby obtaining tenant driving abnormal features. The abnormal driving characteristics of the tenants include the frequency of illegal lane changes, the frequency of running red lights, the frequency of fatigued driving, and the frequency of speeding; the abnormal credit characteristics of the tenants are obtained by extracting features from the preprocessed tenant credit data using LSTM. The tenant credit anomaly features include the frequency of delayed vehicle return, the frequency of delayed payment, and the frequency of payment failure; the tenant behavior anomaly analysis layer obtains the tenant driving anomaly coefficient based on the tenant driving anomaly features; the tenant credit anomaly coefficient is obtained based on the tenant credit anomaly features, and then the tenant behavior anomaly coefficient is obtained; the tenant behavior anomaly coefficient output layer is used to output the tenant behavior anomaly coefficient.

4. The data center abnormal behavior analysis system based on artificial intelligence according to claim 1, characterized in that: The tenant behavior anomaly coefficient is: Using the tenant behavior anomaly impact coefficient as a weight, first calculate the weighted sum of the tenant driving anomaly coefficient and the tenant credit anomaly coefficient, and then use this sum as an exponential function. The input is used to obtain the tenant behavior anomaly coefficient.

5. The data center abnormal behavior analysis system based on artificial intelligence according to claim 1, characterized in that: The vehicle status anomaly identification model includes a vehicle status data input layer, a vehicle status data preprocessing layer, a vehicle status anomaly feature extraction layer, a vehicle status anomaly feature analysis layer, and a vehicle status anomaly coefficient output layer. The vehicle status data input layer is used to input the vehicle status data into the vehicle status anomaly identification model. The vehicle status data preprocessing layer is used to clean the vehicle status data. The vehicle status anomaly feature extraction layer is used to analyze the preprocessed vehicle status data according to CNN to obtain vehicle status anomaly features. The abnormal vehicle status characteristics include the frequency of false alarms, the frequency of abnormal vehicle locations, and the frequency of abnormal vehicle occupancy. The vehicle state anomaly feature analysis layer is used to obtain the vehicle state anomaly coefficient based on the vehicle state anomaly features; the vehicle state anomaly coefficient output layer is used to output the vehicle state anomaly coefficient.

6. The data center abnormal behavior analysis system based on artificial intelligence according to claim 1, characterized in that: The overall car rental anomaly coefficient is: The comprehensive car rental anomaly coefficient is obtained by weighting the tenant behavior anomaly coefficient, vehicle status anomaly coefficient, and platform operation anomaly coefficient by using the car rental anomaly impact coefficient as a weight.

Images