Method and system for pre-judging fatigue driving of online car-hailing driver

Abstract

The invention discloses an online car-hailing driver fatigue driving pre-judgment method and system. The invention relates to the technical field of online car hailing. And collecting driver departure data O and driver order receiving data M which are arranged according to a time sequence. O = [O1, O2,..., ON]; m = [M1, M2,..., MN]; o1, O2,..., ON and M1, M2,..., MN are historical data of driver departure data O and driver order receiving data M respectively. Executing exponential smoothing according to the trend and the seasonality of the driver departure data O and the driver order receiving data M in the time sequence; by adopting the Holt-Wittes seasonal model, the method can accurately capture the periodicity and tendency characteristics of the driving and order receiving behaviors of the driver, and significantly improves the prediction accuracy. The automatic data processing and prediction process greatly shortens the prediction period, improves the operation efficiency, and enables the platform to respond to the potential risk of fatigue driving of the driver more quickly.

Description

Technical Field

[0001] The present invention relates to the technical field of online car-hailing, specifically to driver behavior prediction, and particularly to a method and system for predicting fatigue driving of online car-hailing drivers. Background Art

[0002] As an important branch of the biometric field, face recognition identity verification technology has demonstrated its high efficiency and accuracy in multiple security verification scenarios. On this basis, regarding the problem of identifying fatigue driving of online car-hailing drivers, the academic and industrial communities have explored various technical paths, aiming to improve road transportation safety through intelligent means. These technologies mainly cover detection methods based on motor vehicle behavior characteristics, detection methods based on driver operation behavior, and detection methods based on driver facial characteristics, etc. Typical traditional technologies include:

[0003] (1) CN202310247655.8 discloses a driving style evaluation method, device, medium, and equipment based on an in-vehicle display screen (publication date: July 18, 2023): By analyzing the dynamic behavior characteristics of the vehicle during driving, such as lane-keeping ability, speed change pattern, acceleration fluctuation, etc., to indirectly judge the driver's fatigue state. Abnormal fluctuations in vehicle behavior are often closely related to the driver's distracted attention or slow reaction, thus serving as an indication signal for fatigue driving.

[0004] (2) CN202410101754.X discloses an automotive network security system and method based on driver behavior analysis, a detection method based on motor vehicle behavior characteristics (publication date: January 24, 2024): including the steering wheel rotation frequency, pedal usage force and frequency, gear shifting operation, etc. By comparing the operation mode in the normal driving state with that in fatigue driving, the fatigue degree of the driver can be identified. This method requires the system to accurately capture and analyze the subtle operation changes of the driver.

[0005] (3) CN202411644681.5 discloses a method for detecting fatigue driving of logistics transport drivers (publication date: December 20, 2024): Using advanced image processing and machine learning technologies, this method directly evaluates the driver's fatigue state by analyzing the driver's facial expressions, eye activities (such as blinking frequency, eyelid closing time), head posture, etc. The integration of face recognition technology enables the system to monitor the driver's physiological state in real time without disturbing the driver, providing a direct basis for fatigue driving warning.

[0006] Although certain progress has been made in the above technologies in terms of theory and application, there are significant differences in age, gender, driving experience, physical fitness, etc. among drivers. These individual differences lead to the diversity of fatigue manifestations, making it difficult for a unified standard fatigue recognition algorithm to adapt to all drivers, and subjective factors have a greater impact on the recognition results. Most technologies rely on specific external devices, such as infrared cameras for monitoring eye activities. The installation and maintenance of these devices increase the complexity and cost of the system, and may lead to a decrease in recognition accuracy due to equipment failures or improper calibration.

[0007] As a product of the sharing economy, the configuration levels of intelligent auxiliary hardware devices in online car-hailing vehicles vary widely. Some vehicles lack necessary sensors or computing units, resulting in the inability to effectively deploy the above technologies on these vehicles, which limits the generality of the technologies. Whether it is a detection method based on vehicle behavior, driver operation, or facial features, its implementation often needs to rely on the device hardware of the motor vehicle itself or additional installed devices, which not only increases the purchase or modification cost of the vehicle, but also constitutes an obstacle to large-scale promotion.

[0008] Therefore, the present invention proposes a method and system for predicting fatigue driving of online car-hailing drivers. Summary of the Invention

[0009] In view of this, the present invention hopes to provide a method and system for predicting fatigue driving of online car-hailing drivers to solve or alleviate the technical problems existing in the prior art, namely, significant individual differences among drivers, dependence on external devices, large differences in hardware conditions, and high implementation costs. The technical solution of the present invention is realized as follows:

[0010] In the first aspect, a method for predicting fatigue driving of online car-hailing drivers:

[0011] (1) Overview:

[0012] The present invention aims to deeply analyze and predict the historical data of a driver's trips and order pickups through the Holt-Winters seasonal model to identify the driver's behavior patterns. Specifically, the present invention first collects and arranges the driver's trip and order pickup data in chronological order, and then uses the Holt-Winters model to perform exponential smoothing on these data to estimate the level, trend, and seasonal components, thereby generating accurate prediction values. Then, the present invention determines whether there is convergence between the predicted values of the driver's trips and order pickups at a certain future moment as the basis for whether the driver is in a fatigued state. If the prediction result shows that the driver may be fatigued, the platform can accordingly adjust the order acceptance limit or require the driver to rest, thereby reducing the safety risks brought by fatigue driving.

[0013] (2) Technical solution of the present invention:

[0014] To achieve the above technical goals, the present invention selects the following operating steps to execute.

[0015] 2.1 Step S1, data preparation:

[0016] Collect the driver's vehicle departure data O and driver's order receiving data M and arrange them in chronological order.

[0017] O = [O1, O2,..., ON];

[0018] M = [M1, M2,..., MN];

[0019] Among them, O1, O2,..., ON and M1, M2,..., MN are the historical data of the driver's vehicle departure data O and driver's order receiving data M respectively.

[0020] 2.2 Step S2, perform exponential smoothing:

[0021] Perform exponential smoothing according to the trends and seasonality in the time series of the driver's vehicle departure data O and driver's order receiving data M.

[0022] 3.2.1 Training method:

[0023] Use the Holt-Winters Seasonal Model. And estimate the smoothing parameter α of the level component, the smoothing parameter β of the trend component, and the smoothing parameter γ of the seasonal component. Estimate by minimizing the mean squared error MSE, and use the historical data of the driver's vehicle departure data O and driver's order receiving data M for training:

[0024] (1) Level calculation:

[0025] Among them, L t is the level component at time t, y t is the observation value at time t, S t-s is the seasonal component at time t-s (s is the length of the season cycle), α is the smoothing parameter of the level component (0 ≤ α ≤ 1), L t-1 and T t-1 are the level component and trend component of the previous time point respectively.

[0026] (2) Trend calculation: T t = β(L t -L t-1 )+(1-β)T t-1 ;

[0027] Among them, T t is the trend component at time t, and β is the smoothing parameter of the trend component (0 ≤ β ≤ 1).

[0028] (3) Calculate the seasonal variation:

[0029] Among them, S t is the seasonal component at time t, and γ is the smoothing parameter of the seasonal component (0 ≤ γ ≤ 1).

[0030] (4) Use the mean square error MSE to guide the loss:

[0031] Among them, n is the number of observations, is the predicted value at time t, calculated by the model:

[0032] Use the historical data of the time series of the driver's vehicle departure data O and the driver's order receiving data M for training. When the preset number of training times is reached, find the smoothing parameter α of the level component, the smoothing parameter β of the trend component, and the smoothing parameter γ of the seasonal component when the value of the mean square error MSE is the smallest as the optimal model parameters.

[0033] 3.2.2 Step S200, perform exponential smoothing:

[0034] Apply the Holt-Winters seasonal model to the driver's vehicle departure data O and obtain their respective predicted values.

[0035] (1) Apply the Holt-Winters seasonal model to the driver's vehicle departure data O:

[0036] (1.1) The level of the driver's vehicle departure data O:

[0037] (1.2) The trend of the driver's vehicle departure data O:

[0038] (1.3) The seasonality of the driver's vehicle departure data O:

[0039] (1.4) The prediction of the driver's vehicle departure data O:

[0040] Among them, is the vehicle departure data of the driver at t + h, and respectively represent the level, trend, and seasonal components of the driver's vehicle departure data O at time t. α O , β O and γ O are the smoothing parameters corresponding to the driver's vehicle departure data O.

[0041] (2) Apply the Holt-Winters seasonal model to the driver's order-taking data M:

[0042] (2.1) The level of the driver's order-taking data M:

[0043] (2.2) The trend of the driver's order-taking data M:

[0044] (2.3) The seasonality of the driver's order-taking data M:

[0045] (2.4) The prediction of the driver's order-taking data M:

[0046] Wherein, is the order-taking data of the driver at t + h, and respectively represent the level, trend and seasonal component of the driver's order-taking data M at time t. α M , β M and γ M are the smoothing parameters of the driver's order-taking data M.

[0047] 2.3 Step S3, make a prediction:

[0048] Since the platform will count the online driving duration of the driver for each working day, and record and count the driving duration and driving mileage of each order. According to the driving data of the driver at t + h and the order-taking data at t + h to determine whether there is convergence. If there is convergence, it proves that the driver has a driving behavior and an order-taking behavior at t + h, indicating that the driver is not in a fatigued state at that time. On the contrary, if there is no convergence, it proves that the driver either has no driving behavior or no order-taking behavior at t + h, indicating that the driver is in a fatigued state.

[0049] Because the driving data and the order-taking data at t + h are of the same scalar, a threshold T is set for judgment. If the driving data and the order-taking data at t + h both exceed the threshold T, it proves that there is convergence; if either one or both of the driving data and the order-taking data at t + h do not exceed the threshold T, it proves that there is no convergence.

[0050] It is possible to judge whether the driver stops to rest according to the reported location on the driver side, and lift the order-taking restriction after reaching the specified time.

[0051] (3) Mechanism for Solving Technical Problems:

[0052] 3.1 Significantly addressing individual differences among drivers:

[0053] By collecting the driving and order-taking historical data of each driver and using the Holt-Winters seasonal model for personalized prediction. This model can capture the unique behavior patterns of each driver, including the level, trend, and seasonal components, thus accurately predicting the driving and order-taking behaviors of the driver. This method avoids the problem of inaccurate prediction caused by ignoring individual differences among drivers in traditional technologies.

[0054] 3.2 Solving the dependence on external devices:

[0055] The present invention completely relies on the historical data of drivers' driving and order-taking for prediction, without the need for external devices to collect data in real time. This reduces the dependence on external devices, improves the flexibility and reliability of the system. At the same time, it also reduces the risks brought by external device failures or inaccurate data.

[0056] 3.3 Solving large differences in hardware conditions:

[0057] The Holt-Winters seasonal model has good adaptability and can process data collected under different hardware conditions. No matter how the hardware conditions change, as long as the data formats are consistent, the model can effectively make predictions. In addition, the present invention has low requirements for hardware resources, does not require high-performance computing devices, and reduces the hardware cost.

[0058] 3.4 Solving high implementation costs:

[0059] Using the exponential smoothing algorithm and the Holt-Winters seasonal model for data processing and prediction reduces the computational complexity and implementation costs. At the same time, through the reuse of historical data and the self-adaptability of the model, the dependence on new data and the retraining cost of the model are reduced. Through the automated data collection, processing, and prediction processes, the need for human intervention is reduced, and the labor cost is lowered. In addition, the intuitiveness and ease of use of the prediction results also reduce the usage threshold and training costs for platform operators.

[0060] Second, the fatigue driving pre-judgment system for online car-hailing drivers:

[0061] As Figure 4 shown, this system is used to implement the fatigue driving pre-judgment method for online car-hailing drivers described above, and it includes:

[0062] (1) Data acquisition layer: Responsible for collecting the driving and order-taking historical data of drivers. These data can be obtained through the database or log system of the online car-hailing platform.

[0063] It includes a data collection module: responsible for docking with the database or log system of the online car-hailing platform, and collecting the driving and order-taking historical data of drivers in real time or regularly. Transmit the collected data to the data processing module.

[0064] (2) Data processing layer: Clean, organize and analyze the collected raw data, and extract the feature information useful for prediction. It includes a data processing module: Receive the data transmitted by the data collection module, and perform cleaning, organizing and analyzing. Extract the feature information useful for prediction, such as the driving frequency, order-taking quantity and driving mileage of the driver. Transmit the processed data to the model prediction module.

[0065] (3) Model prediction layer: Predict the processed data to generate the prediction results of the driver's future driving and order-taking behaviors. It includes a model prediction module: Use the Holt-Winters seasonal model algorithm to predict the data transmitted by the data processing module. Generate the prediction results of the driver's future driving and order-taking behaviors. Transmit the prediction results to the decision-making and feedback module.

[0066] (4) Decision-making and feedback layer: Formulate decision-making measures according to the prediction results. It includes a decision-making and feedback module: According to the prediction results transmitted by the model prediction module, formulate corresponding decision-making measures. If it is predicted that the driver is fatigued while driving, adjust the order-taking limit or remind the driver to rest. Feed back the decision-making information to the driver and platform operators so that corresponding measures can be taken in a timely manner.

[0067] Compared with the prior art, the beneficial effects of the present invention are:

[0068] First, improve the prediction accuracy and efficiency: By adopting the Holt-Winters seasonal model, the present invention can accurately capture the periodic and trend characteristics of the driver's driving and order-taking behaviors, and significantly improve the prediction accuracy. The automated data processing and prediction process greatly shortens the prediction cycle, improves the operation efficiency, and enables the platform to respond more quickly to the potential risks of driver fatigue while driving.

[0069] Second, strengthen the driver safety management: The present invention can monitor the driver's working status in real time, discover and warn of fatigue driving behaviors in a timely manner, effectively prevent traffic accidents from occurring, and ensure the lives and safety of drivers and passengers. By adjusting the order-taking quantity, arranging rest time and other measures, it helps drivers reasonably arrange work and rest, promotes the physical and mental health of drivers, and reduces the safety hazards caused by fatigue driving.

[0070] Third, optimize the resource allocation: The accurate prediction results of the present invention enable the platform to allocate driver resources more reasonably, avoid the situation of driver shortage or surplus during peak hours, and improve the resource utilization efficiency. By predicting the driver's driving behavior, the platform can more effectively plan vehicle dispatching, reduce the empty driving rate, and lower the operation cost.

[0071] IV. Improving User Experience: The accurate prediction of the order-taking time in the present invention improves the travel experience of passengers, reduces the waiting time, and enhances the trust and satisfaction of passengers towards the platform. When drivers arrange their work and rest reasonably, they can provide more high-quality and efficient services, further enhancing the travel experience of passengers. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following-described drawings are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0073] Figure 1 It is a schematic flowchart of the method of the present invention;

[0074] Figure 2 It is a schematic diagram of the method execution of the present invention;

[0075] Figure 3 It is a schematic flowchart of the model calculation of the present invention;

[0076] Figure 4 It is a schematic diagram of the system composition of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0077] In order to make the above-mentioned objects, features, and advantages of the present invention more obvious and understandable, the following will provide a detailed description of the specific embodiments of the present invention with reference to the drawings. Many specific details are set forth in the following description in order to fully understand the present invention. However, the present invention can be implemented in many other ways different from those described herein. Those skilled in the art can make similar improvements without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below;

[0078] It should be noted that the embodiments in this specification are described in a progressive manner, and the key points of each embodiment are the differences from other embodiments. The same or similar parts among the embodiments can be referred to each other. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the description of the method part.

[0079] Explanation of Related Terms:

[0080] (1) Driver's vehicle departure data O: A historical data sequence recording the driver's vehicle departure behavior.

[0081] (2) Driver's order-taking data M: A historical data sequence recording the driver's order-taking behavior.

[0082] (3) Holt-Winters Seasonal Model: A model for time series prediction that takes into account the level, trend, and seasonal components.

[0083] (4) Level: The baseline value or average in time series data.

[0084] (5) Trend: The long-term direction or pattern that changes over time in time series data.

[0085] (6) Seasonality: The periodic pattern that appears due to seasonal changes in time series data.

[0086] (7) Smoothing parameter α: A parameter used to control the weight of the level component in prediction.

[0087] (8) Smoothing parameter β: A parameter used to control the weight of the trend component in prediction.

[0088] (9) Smoothing parameter γ: A parameter used to control the weight of the seasonal component in prediction.

[0089] (10) The driving data of the driver at t+h and the order receiving data at t+h: The driving and order receiving behavior data of the driver at a future moment to be predicted.

[0090] (11) Convergence: The consistency or similarity shown by the predicted data of the driver's driving and order receiving at a certain moment.

[0091] Example 1: As Figures 1 to 4 shown, this example discloses a method for predicting fatigue driving of online car-hailing drivers; assume that the online car-hailing platform needs to daily manage and operate the driving and order receiving behaviors of drivers to predict whether the driver is in a fatigue driving state. By collecting the driving data and order receiving data of the driver and using the Holt-Winters seasonal model for exponential smoothing prediction, the fatigue state of the driver is further judged.

[0092] In this example, regarding step S1, data preparation: Collect the driving data O of the driver and the order receiving data M arranged in chronological order.

[0093] O = [O1, O2,..., ON];

[0094] M = [M1, M2,..., MN];

[0095] where O1, O2,..., ON and M1, M2,..., MN are the historical data of the driving data O of the driver and the order receiving data M respectively.

[0096] Further:

[0097] (1) Driver's vehicle departure data O: including the start time, end time, and total vehicle departure duration of the driver each day.

[0098] (2) Driver's order receiving data M: including the number of orders received by the driver each day, the vehicle departure duration of each order, and the driving mileage.

[0099] Among them, O1 represents the vehicle departure data on the first day, and ON represents the vehicle departure data on the Nth day; similarly, M = [M1, M2,..., MN].

[0100] In this embodiment, regarding step S2, exponential smoothing is performed: the Holt-Winters seasonal model is used to train the driver's vehicle departure data O and order receiving data M. Estimate the smoothing parameter α of the level component, the smoothing parameter β of the trend component, and the smoothing parameter γ of the seasonal component. Find the optimal model parameters by minimizing the mean square error MSE.

[0101] Specifically, first, the following training method needs to be performed on this model: use the Holt-Winters Seasonal Model. And estimate its smoothing parameter α of the level component, the smoothing parameter β of the trend component, and the smoothing parameter γ of the seasonal component, and estimate by minimizing the mean square error MSE, and use the historical data of the driver's vehicle departure data O and driver's order receiving data M for training:

[0102] (1) Level calculation:

[0103] Among them, L t is the level component at time t, y t is the observed value at time t, S t-s is the seasonal component at time t - s (s is the length of the season cycle), α is the smoothing parameter of the level component (0 ≤ α ≤ 1), L t-1 and T t-1 are the level component and trend component of the previous time point respectively.

[0104] (2) Trend calculation: T t = β(L t - L t-1 )+(1 - β)T t-1 ;

[0105] Among them, T t is the trend component at time t, and β is the smoothing parameter of the trend component (0 ≤ β ≤ 1).

[0106] (3) Calculate seasonal variation:

[0107] Among them, S tis the seasonal component at time t, and γ is the smoothing parameter of the seasonal component (0 ≤ γ ≤ 1).

[0108] (4) The mean squared error MSE is used to guide the loss:

[0109] where n is the number of observations, is the predicted value at time t, calculated by the model:

[0110]

[0111] Use the historical data of the time series of driver departure data O and driver order receiving data M for training. When the preset number of training times is reached, find the smoothing parameter α of the level component, the smoothing parameter β of the trend component, and the smoothing parameter γ of the seasonal component when the value of the mean squared error MSE is the smallest as the optimal model parameters.

[0112] It can be understood that through the adaptive adjustment of the smoothing parameters (α, β, γ) in the model, the model can adapt to the departure and order receiving behavior patterns of different drivers. This solution mainly relies on the existing data records of the platform and does not rely on external devices additionally, reducing the implementation cost. Since the data is collected and processed uniformly by the platform, it is not affected by the differences in the hardware conditions of the driver side.

[0113] Specifically, apply the Holt-Winters seasonal model to the driver departure data O and obtain their respective predicted values.

[0114] (1) Apply the Holt-Winters seasonal model to the driver departure data O:

[0115] (1.1) The level of the driver departure data O:

[0116] (1.2) The trend of the driver departure data O:

[0117] (1.3) The seasonality of the driver departure data O:

[0118] (1.4) The prediction of the driver departure data O:

[0119] where, is the departure data of the driver at t + h, and represent the level, trend, and seasonal components of the driver departure data O at time t, respectively. α O 、β O and γ OIt is the smoothing parameter corresponding to the driver’s dispatch data O.

[0120] Specifically, the Holt-Winters seasonal model is applied to the driver order data M:

[0121] (2.1) The level of driver order data M:

[0122] (2.2) Trend of driver order data M:

[0123] (2.3) Seasonality of driver order data M:

[0124] (2.4) Prediction of driver order data M:

[0125] in, It is the driver's order data at t+h. and They represent the level, trend and seasonal components of the driver order data M at time t. M , β M and γ M is the smoothing parameter of the driver’s order data M.

[0126] In this embodiment, regarding step S3, prediction is performed: because the platform will count the online time of the driver every working day, and record and count the time and mileage of each order. And the order data at t+h Determine whether there is convergence. If there is convergence, it proves that the driver has dispatched the vehicle and accepted the order at t+h, which proves that the driver was not fatigued at that time. On the contrary, if there is no convergence, it proves that the driver either did not dispatch the vehicle or did not accept the order at t+h, which proves that the driver was fatigued.

[0127] Because the vehicle data And the order data at t+h is the same scalar, so a threshold T is set to judge. And the order data at t+h If the vehicle data exceeds the threshold T, it proves that there is convergence; And the order data at t+h If either or both of them do not exceed the threshold T, there is no convergence.

[0128] The system can determine whether the driver should stop for a rest based on the reported location on the driver's side, and lift the order restriction after the specified time.

[0129] It is understandable that by predicting the driver's fatigue state, timely reminding the driver to rest, and reducing the risk of traffic accidents caused by fatigue driving. Automatically judging the driver's fatigue state reduces manual intervention and improves the platform operation efficiency. Reasonably arranging the driver's rest time to avoid overwork enhances the driver's satisfaction and loyalty. Without relying on external devices and using existing data for prediction reduces the implementation and maintenance costs.

[0130] Embodiment 2: On the basis of Embodiment 1, this embodiment will further provide its specific Python execution program:

[0131]

[0132]

[0133]

[0134]

[0135] In the above program: Collect the driver's vehicle departure data and order receiving data and arrange them in chronological order. Use the Pandas library to organize the data into a DataFrame format for easy processing. Use the ExponentialSmoothing class in the statsmodels library to implement the Holt-Winters seasonal model. Train the models for the vehicle departure data (such as the total vehicle departure duration) and the order receiving data (such as the number of orders received) respectively. Estimate the smoothing parameters of the level, trend, and seasonal components.

[0136] Use the trained models to predict the vehicle departure data and order receiving data for the next day. Set a threshold T for judging the convergence of the predicted data.

[0137] Compare the convergence of the predicted data. If there is convergence between the predicted values (i.e., the difference is less than the threshold T), it is considered that the driver is in a non-fatigued state. If there is no convergence, it is considered that the driver may be in a fatigued state. Finally, output the prediction results and the fatigue state judgment.

[0138] All of the above embodiments only express the implementation manners of the relevant actual applications of the present invention. The descriptions are relatively specific and detailed, but should not be construed as a limitation on the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several deformations and improvements can still be made, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the present invention patent shall be subject to the appended claims.

[0139] For those skilled in the art, it can be further realized that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of the two. To clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described according to functions in the above description. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods for each specific application to implement the described functions, but such implementation should not be considered to exceed the scope of the present invention.

[0140] At the same time, those skilled in the art can understand that all or part of the processes of implementing the methods of all the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to the memory, storage, database, or other media provided in the present application and used in the embodiments can include non-volatile and / or volatile memories. Non-volatile memories can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memories can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

Claims

1. A method for predicting fatigue driving of online car-hailing drivers, characterized in that, Including: S1, driver's vehicle operation data O and driver's order receiving data M collected and arranged in chronological order; S2, performing exponential smoothing according to the trends and seasonality of the driver's vehicle operation data O and the driver's order receiving data M in the time series; S3. According to the driver's trip data at t+h and the order receiving data at t+h judge whether there is convergence; if there is convergence, it proves that the driver has a trip behavior and an order receiving behavior at t+h, indicating that the driver is not in a fatigued state at that time; conversely, if there is no convergence, it proves that the driver is in a fatigued state at t+h.

2. The prediction method according to claim 1, characterized in that: In the said S1: O = [O1, O2,..., ON]; M = [M1, M2,..., MN]; wherein, O1, O2,..., ON and M1, M2,..., MN are respectively the historical data of the driver's vehicle operation data O and the driver's order receiving data M.

3. The prediction method according to claim 2, wherein: In the said S2, including: Applying the Holt-Winters seasonal model to the driver's vehicle operation data O and obtaining their respective predicted values: Level of driver's vehicle departure data O: Trend of driver's vehicle departure data O: Seasonality of Driver Trip Data O: Prediction of driver's vehicle operation data O: Among them, is the vehicle dispatch data of the driver at t + h, and respectively represent the level, trend and seasonal components of the driver's vehicle dispatch data O at time t; α O , β O and γ O are the smoothing parameters corresponding to the driver's vehicle dispatch data O.

4. The pre-judgment method according to claim 3, wherein: In the said S2, including: Level of driver's order receiving data M: Trend of driver's order receiving data M: Seasonality of driver order acceptance data M: Prediction of driver order acceptance data M: Among them, is the order receiving data of the driver at t + h, and respectively represent the level, trend and seasonal components of the driver's order receiving data M at time t; α M , β M and γ M are the smoothing parameters of the driver's order receiving data M.

5. The prediction method according to claim 4, wherein: The training method of the Holt-Winters seasonal model used in step S2 includes: Horizontal calculation: where L t is the horizontal component at time t, y t is the observed value at time t, S t-s is the seasonal component at time t - s, α is the smoothing parameter of the horizontal component, L t-1 and T t-1 are the horizontal component and the trend component at the previous time point, respectively; Trend calculation: T t = β(L t - L t-1 ) + (1 - β)T t-1 ; where, T t is the trend component at time t, and β is the smoothing parameter of the trend component; calculate the seasonal variation: where S t is the seasonal component at time t, and γ is the smoothing parameter of the seasonal component; Mean Squared Error MSE is used to guide the loss: where n is the number of observations, is the predicted value at time t.

6. The prediction method according to claim 5, wherein: Predicted value Calculated by the model: Using the historical data of the time series of the driver's vehicle operation data O and the driver's order receiving data M for training. When the preset number of training times is reached, find the smoothing parameter α of the level component, the smoothing parameter β of the trend component, and the smoothing parameter γ of the seasonal component when the value of the mean square error MSE is the smallest, as the optimal model parameters.

7. The prediction method according to any one of claims 1 to 6, characterized in that: In the step S3, if the vehicle departure data and the order receiving data at t+h both exceed the threshold value T, it proves that there is convergence; if either one or both of the vehicle departure data and the order receiving data at t+h do not exceed the threshold value T, it proves that there is no convergence.

8. The pre-judgment method according to claim 7, wherein: Make a judgment on whether the driver stops to rest according to the reported location of the driver's end, and lift the order receiving restriction after reaching the specified time.

9. A system for implementing the pre-judgment method according to any one of claims 1 to 8, characterized in that, The said system includes: Data acquisition layer: responsible for collecting the historical data of the driver's vehicle operation and order receiving; Data processing layer: cleaning, sorting and analyzing the collected raw data; Model prediction layer: predicting the processed data to generate the prediction results of the driver's future vehicle operation and order receiving behaviors; Decision-making and feedback layer: formulating decision-making measures according to the prediction results.

10. The system according to claim 9, wherein: The said system includes: Data acquisition module: responsible for docking with the database or log system of the online car-hailing platform to collect the historical data of the driver's vehicle operation and order receiving in real time or regularly; Data processing module: receiving the data transmitted by the data acquisition module, cleaning, sorting and analyzing it; transmitting the processed data to the model prediction module; Model prediction module: using the Holt-Winters seasonal model algorithm to predict the data transmitted by the data processing module; generating the prediction results of the driver's future vehicle operation and order receiving behaviors; transmitting the prediction results to the decision-making and feedback module; Decision-making and feedback module: formulating corresponding decision-making measures according to the prediction results transmitted by the model prediction module; predicting that the driver is fatigued, then adjusting the order receiving restriction or reminding the driver to rest; feeding back the decision-making information to the driver and the platform operators.

Images