A method for identifying mileage fraud at turnaround points in online ride-hailing services based on a multi-dimensional model

Abstract

The present invention relates to a method for identifying mileage cheating at turnaround points of online ride-hailing vehicles based on a multidimensional model, and specifically relates to the field of traffic analysis. The method collects driving data and trajectory data of online ride-hailing vehicles and checks and corrects the driving data, reduces the interference of abnormal driving data on subsequent analysis, improves the accuracy and credibility of driving data, uses a decision tree to construct a driving behavior model to classify actual driving behavior, establishes a driver credit point system, and adds or subtracts driver credit points based on passenger evaluation, number of completed orders, and passenger complaints. The method collects data on the city where the online ride-hailing vehicles are located, and uses a decision tree to construct a city point quality model to classify weather, tunnels, and mountains. The driving behavior model and the city point quality model are combined to evaluate the expected quality score of each location point in the expected behavior of the trajectory data, which helps to identify and distinguish abnormal situations of different degrees and promote drivers to continuously provide high-quality online ride-hailing services.

Description

Technical Field

[0001] The present invention relates to the field of traffic analysis, and more specifically, to a method for identifying mileage cheating at turnaround points in online ride-hailing vehicles based on a multi-dimensional model. Background Art

[0002] With the rapid development of the online car-hailing market, the problem of cheating at turn-around points has gradually become prominent. Some drivers have caused economic losses to the platform and users and disrupted market order by tampering with GPS positioning and false driving.

[0003] Currently, cheating is determined by calculating the direction, speed, and distance at the turning points in the driving trajectory of online ride-hailing vehicles to see if they exceed the system configuration threshold. The factors of direction, speed, and distance are too simple and have a high probability of misjudgment, causing losses to drivers and passengers. There is no consideration of the turning point phenomenon caused by drifting or missing points due to environmental geography. Summary of the Invention

[0004] In response to the technical problems existing in the prior art, the present invention provides a method for identifying mileage cheating at turnaround points of online ride-hailing vehicles based on a multi-dimensional model to solve the problems raised in the above-mentioned background technology.

[0005] The present invention solves the above-mentioned technical problem with the following technical solution: a method for identifying mileage fraud at turnaround points in online ride-hailing vehicles based on a multi-dimensional model, comprising the following steps:

[0006] S101. Collect driving data and trajectory data from online ride-hailing services and perform error checking and correction on the driving data. Add a sliding window to extract the average eigenvalue of the driving data. Use a decision tree to build a driving behavior model to classify actual driving behavior.

[0007] S102. Establish a driver credit score system and add or subtract driver credit scores based on passenger reviews, number of completed orders, and passenger complaints;

[0008] S103. Collect data on the cities where ride-hailing services are located, divide the city data into regular 5-square-meter grids, count the number of points that meet the requirements, use Euclidean methods to determine the distance between the location point and the nearest road network, and construct a city point quality model using a decision tree to classify weather, tunnels, and mountains.

[0009] S104. Based on the classification of actual driving behavior in the driving behavior model, the expected behavior of the trajectory data is obtained. The expected quality score of each location point in the expected behavior of the trajectory data is evaluated using the city point quality model. The quality scores of the actual driving trajectories are compared, and the difference is determined as an outlier. The credibility value of the actual driving trajectory is used as an influencing factor to recalculate the driver's credit score.

[0010] In a preferred embodiment, in S101, the driving data and trajectory data of the online car-hailing service are collected, including the distance between adjacent driving points, driving speed, driving direction, and GPS signal strength during driving, and the error values ​​in the driving data of the online car-hailing service are checked. For the driving speed and driving direction, it is checked whether they exceed a reasonable range. If they exceed a reasonable range, the data is deleted. For the distance between adjacent driving points and GPS signal strength during driving, it is checked whether there are errors and inconsistent information. If there are errors and inconsistent information, the data is deleted, and preparations are made to fill the missing values ​​in the preliminary driving data items.

[0011] Furthermore, a sliding window is added to apply a fixed-size window of 5 minutes to the driving data, and the sliding is performed with a fixed step size of half the window length. The fine-grainedness of the driving data is controlled by adjusting the window size and step size to capture short-term changes in driving behavior. The sliding window is used to extract the average characteristic values ​​of the distance between adjacent driving points, driving speed, driving direction, and GPS signal strength during the driving process.

[0012] Furthermore, a driving behavior model is constructed using a decision tree to classify actual driving behavior. The average characteristic value of the distance between adjacent driving points, driving speed, driving direction, and GPS signal strength during driving is selected as the root node of the decision tree, and the driving data is divided into different driving subsets. For each non-leaf node, the best feature among the average characteristic value of the distance between adjacent driving points, driving speed, driving direction, and GPS signal strength during driving is selected for splitting. The smaller the Gini index of the non-leaf node, the lower the impurity of the node, and the more consistent the driving data category represented by the feature. The current node is split into multiple child nodes based on the selected best feature, where each child node corresponds to a different driving subset under the feature value. The selection and splitting steps are repeated until the driving data represented by the corresponding feature of the node belongs to the same category and is labeled with the category, including normal driving, speeding, sudden braking, and abnormal turning.

[0013] In a preferred embodiment, in S102, a driver credit score system is established. The default value of the credit score of a newly registered driver is 50 points. For the passenger evaluation dimension, points are added or subtracted according to the star rating of the passenger evaluation. If a passenger gives a 5-star evaluation, the driver's credit score is added by 0.5 points, if a passenger gives a 4-star evaluation, the driver's credit score is added by 0.2 points, if a passenger gives a 3-star evaluation, the driver's credit score is added by 0 points, if a passenger gives a 2-star evaluation, the driver's credit score is subtracted by 0.3 points, and if a passenger gives a 1-star evaluation, the driver's credit score is subtracted by 0.5 points. For the dimension of the number of completed orders, points are calculated based on the number of orders completed by the driver without fault. The driver's credit score is added by 3 points for every 100 orders completed. For the dimension of passenger complaints, if a customer complains and it is confirmed that the driver is responsible, the driver's credit score is subtracted by 2 points. When the cumulative number of customer complaints and confirmations that the driver is responsible is between 3 and 5 times, the driver's credit score is subtracted by 3 points each time. When the cumulative number of customer complaints and confirmations that the driver is responsible is between 3 and 5 times, the driver's credit score is subtracted by 4 points each time. When the cumulative number of customer complaints and confirmations that the driver is responsible exceeds 10 times, the driver's credit score is subtracted by 5 points each time.

[0014] In a preferred embodiment, in S103, data on the city where the online ride-hailing service is located, including weather, tunnels, and mountains, is collected. Combined with the GPS signal strength during driving, the satellite map is connected to check for erroneous values ​​in the city data and replace them. The city data is divided into regular grid areas of 5 square meters in size. Within each grid, the number of points that meet the conditions based on weather, tunnels, and mountains is counted. For each location point, a spatial index structure is used to accelerate the nearest search, and the distance between the location point and the nearest road network is obtained using Euclidean method. The specific Euclidean formula is as follows:

[0015]

[0016] Among them, (X1, Y1) represents the coordinates of the location point in the grid area, (X2, Y2) represents the coordinates of the nearest road network in the grid area, S represents the distance between the location point and the nearest road network, and the distance between the location point and the nearest road network and the number of points are used as feature values.

[0017] Furthermore, a decision tree is used to construct a city point quality model to classify weather, tunnels and mountains. The characteristic values ​​represented by the distance between the location point and the nearest road network and the number of points are selected as the root node of the decision tree, and the city data is divided into different city subsets. For each non-leaf node, the best feature of the distance between the location point and the nearest road network and the number of points is selected for splitting. The smaller the Gini index of the non-leaf node, the lower the impurity of the node, and the more consistent the city data category represented by the feature. The current node is split into multiple child nodes according to the selected best feature, where each child node corresponds to a different city subset under the feature value. The selection and splitting steps are repeated until the city data represented by the node corresponding feature belongs to the same category and is labeled with the category, including weather, tunnels and mountains.

[0018] In a preferred embodiment, in S104, the expected behavior of the trajectory data is obtained according to the classification of the actual driving behavior of the driving behavior model, and the expected quality score of each position point in the expected behavior of the trajectory data is evaluated in combination with the urban point quality model. The expected quality score is compared with the quality score of the actual driving trajectory, and the difference is obtained as the abnormal value, which ranges from [0-10]. The abnormal value range of [0-3] is considered normal, the abnormal value range of [3-5] is considered slightly abnormal, the abnormal value range of [5-8] is considered abnormal, and the abnormal value range of [8-10] is considered seriously abnormal. The credibility value of the actual driving trajectory is evaluated according to the driver's credit score, and the credibility value of the actual driving trajectory is used as an influencing factor to recalculate the driver's credit score. The specific calculation formula is:

[0019]

[0020] Among them, P represents the credibility value of the actual driving trajectory, t represents the abnormal value, e represents the driver's credit score value, r represents the impact factor, the driving trajectory credibility value range is [0-10], among which the driving trajectory credibility value range is [0-2] is considered untrustworthy, among which the driving trajectory credibility value range is [2-5] is considered slightly untrustworthy, among which the driving trajectory credibility value range is [5-8] is considered slightly trustworthy, among which the driving trajectory credibility value range is [8-10] is considered trustworthy. When the driving trajectory credibility value range is [0-2], the driver's credit score is reduced by 2 points, and when the driving trajectory credibility value range is [2-5], the driver's credit score is reduced by 1 point. The driving trajectory in the untrustworthy range is settled and marked, and the marked driving trajectory is submitted to the manual review team. The review result is used as the influencing factor of the driver credit score system and the driver's score is adjusted.

[0021] The beneficial effects of the present invention are: collecting driving data and trajectory data of online ride-hailing vehicles and checking and correcting the driving data, reducing the interference of abnormal driving data on subsequent analysis, improving the accuracy and credibility of driving data, helping to identify temporary behavior patterns by applying sliding windows, promoting drivers to continue to provide high-quality online ride-hailing services by establishing a driver credit points system, using spatial index structures to accelerate the nearest search, and being able to obtain the distance between the location point and the nearest road network more quickly, using decision trees to construct a city point quality model, which can effectively classify and identify different categories in urban data, comparing the expected quality score with the quality score of the actual driving trajectory, helping to identify and distinguish different degrees of abnormal situations, and providing reliability information about the driving trajectory through driving trajectory credibility value evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 Flow chart of the method of the present invention; DETAILED DESCRIPTION

[0023] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0024] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of the technical features indicated. Therefore, a feature specified as "first" or "second" may explicitly or implicitly include one or more of the described features. In the description of this application, "plurality" means two or more, unless otherwise specifically specified.

[0025] In the description of this application, the term "for example" is used to mean "used as an example, illustration or explanation". Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is given to enable any person skilled in the art to implement and use the present invention. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art will recognize that the present invention can be implemented without using these specific details. In other examples, well-known structures and processes will not be elaborated in detail to avoid obscuring the description of the present invention with unnecessary details. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the widest scope consistent with the principles and features disclosed in this application.

[0026] This embodiment provides Figure 1The method for identifying mileage fraud at turnaround points in online ride-hailing services based on a multi-dimensional model specifically includes the following steps:

[0027] S101. Collect driving data and trajectory data from online ride-hailing services and perform error checking and correction on the driving data. Add a sliding window to extract the average eigenvalue of the driving data. Use a decision tree to build a driving behavior model to classify actual driving behavior.

[0028] Furthermore, the driving data and trajectory data of online ride-hailing vehicles are collected, including the distance between adjacent driving points, driving speed, driving direction, and GPS signal strength during driving, and the error values ​​in the driving data of online ride-hailing vehicles are checked. For driving speed and driving direction, check whether they exceed a reasonable range. If they do, delete the data. For the distance between adjacent driving points and GPS signal strength during driving, check whether there are errors and inconsistent information. If there are errors and inconsistent information, delete the data, and prepare preliminary driving data items to fill in the missing values.

[0029] Furthermore, a sliding window is added to apply a fixed-size window of 5 minutes to the driving data, and the sliding is performed with a fixed step size of half the window length. The fine-grainedness of the driving data is controlled by adjusting the window size and step size to capture short-term changes in driving behavior. The sliding window is used to extract the average characteristic values ​​of the distance between adjacent driving points, driving speed, driving direction, and GPS signal strength during the driving process.

[0030] Furthermore, a driving behavior model is constructed using a decision tree to classify actual driving behavior. The average characteristic value of the distance between adjacent driving points, driving speed, driving direction, and GPS signal strength during driving is selected as the root node of the decision tree, and the driving data is divided into different driving subsets. For each non-leaf node, the best feature among the average characteristic value of the distance between adjacent driving points, driving speed, driving direction, and GPS signal strength during driving is selected for splitting. The smaller the Gini index of the non-leaf node, the lower the impurity of the node, and the more consistent the driving data category represented by the feature. The current node is split into multiple child nodes based on the selected best feature, where each child node corresponds to a different driving subset under the feature value. The selection and splitting steps are repeated until the driving data represented by the corresponding feature of the node belongs to the same category and is labeled with the category, including normal driving, speeding, sudden braking, and abnormal turning.

[0031] S102. Establish a driver credit score system and add or subtract driver credit scores based on passenger reviews, number of completed orders, and passenger complaints;

[0032] Furthermore, a driver credit points system is established. The default credit points for newly registered drivers is 50 points. For the passenger evaluation dimension, points are added or subtracted according to the star rating of the passenger evaluation. If a passenger gives a 5-star rating, the driver's credit points are added by 0.5 points; if a passenger gives a 4-star rating, the driver's credit points are added by 0.2 points; if a passenger gives a 3-star rating, the driver's credit points are added by 0 points; if a passenger gives a 2-star rating, the driver's credit points are subtracted by 0.3 points; and if a passenger gives a 1-star rating, the driver's credit points are subtracted by 0.5 points. For the dimension of the number of completed orders, points are calculated based on the number of orders completed by the driver without fault. The driver's credit points are added by 3 points for every 100 orders completed. For the dimension of passenger complaints, the driver's credit points are subtracted by 2 points if the driver is complained by a customer and confirmed to be responsible. When the cumulative number of complaints by customers and confirmed to be the driver's responsibility is between 3 and 5 times, the driver's credit points are subtracted by 3 points each time. When the cumulative number of complaints by customers and confirmed to be the driver's responsibility is between 3 and 5 times, the driver's credit points are subtracted by 4 points each time. When the cumulative number of complaints by customers and confirmed to be the driver's responsibility exceeds 10 times, the driver's credit points are subtracted by 5 points each time.

[0033] S103. Collect data on the cities where ride-hailing services are located, divide the city data into regular 5-square-meter grids, count the number of points that meet the requirements, use Euclidean methods to determine the distance between the location point and the nearest road network, and construct a city point quality model using a decision tree to classify weather, tunnels, and mountains.

[0034] Furthermore, we collected data on the cities where the ride-hailing service is located, including weather, tunnels, and mountains. Combined with the GPS signal strength during driving, we connected to a satellite map to check for erroneous values ​​in the city data and replaced them. We then divided the city data into regular grid areas of 5 square meters. Within each grid, we counted the number of points that met the conditions based on weather, tunnels, and mountains. For each location, we used a spatial index structure to accelerate the nearest neighbor search, and used Euclidean methods to obtain the distance between the location and the nearest road network. The specific Euclidean formula is:

[0035]

[0036] Among them, (X1, Y1) represents the coordinates of the location point in the grid area, (X2, Y2) represents the coordinates of the nearest road network in the grid area, S represents the distance between the location point and the nearest road network, and the distance between the location point and the nearest road network and the number of points are used as feature values.

[0037] Furthermore, a decision tree is used to construct a city point quality model to classify weather, tunnels and mountains. The characteristic values ​​represented by the distance between the location point and the nearest road network and the number of points are selected as the root node of the decision tree, and the city data is divided into different city subsets. For each non-leaf node, the best feature of the distance between the location point and the nearest road network and the number of points is selected for splitting. The smaller the Gini index of the non-leaf node, the lower the impurity of the node, and the more consistent the city data category represented by the feature. The current node is split into multiple child nodes according to the selected best feature, where each child node corresponds to a different city subset under the feature value. The selection and splitting steps are repeated until the city data represented by the node corresponding feature belongs to the same category and is labeled with the category, including weather, tunnels and mountains.

[0038] S104. Based on the classification of actual driving behavior in the driving behavior model, the expected behavior of the trajectory data is obtained. The expected quality score of each location point in the expected behavior of the trajectory data is evaluated using the city point quality model. The quality scores of the actual driving trajectories are compared, and the difference is determined as an outlier. The credibility value of the actual driving trajectory is used as an influencing factor to recalculate the driver's credit score.

[0039] Furthermore, the expected behavior of the trajectory data is obtained based on the classification of the actual driving behavior of the driving behavior model. The expected quality score of each location point in the expected behavior of the trajectory data is evaluated in combination with the urban point quality model. The expected quality score is compared with the quality score of the actual driving trajectory, and the difference is obtained as the abnormal value, which ranges from [0-10]. The abnormal value range of [0-3] is considered normal, the abnormal value range of [3-5] is considered mildly abnormal, the abnormal value range of [5-8] is considered abnormal, and the abnormal value range of [8-10] is considered severely abnormal. The credibility value of the actual driving trajectory is evaluated based on the driver's credit score, and the driver's credit score is recalculated using the credibility value of the actual driving trajectory as an influencing factor. The specific calculation formula is as follows:

[0040]

[0041] Among them, P represents the credibility value of the actual driving trajectory, t represents the abnormal value, e represents the driver's credit score value, r represents the impact factor, the driving trajectory credibility value range is [0-10], among which the driving trajectory credibility value range is [0-2] is considered untrustworthy, among which the driving trajectory credibility value range is [2-5] is considered slightly untrustworthy, among which the driving trajectory credibility value range is [5-8] is considered slightly trustworthy, among which the driving trajectory credibility value range is [8-10] is considered trustworthy. When the driving trajectory credibility value range is [0-2], the driver's credit score is reduced by 2 points, and when the driving trajectory credibility value range is [2-5], the driver's credit score is reduced by 1 point. The driving trajectory in the untrustworthy range is settled and marked, and the marked driving trajectory is submitted to the manual review team. The review result is used as the influencing factor of the driver credit score system and the driver's score is adjusted.

[0042] It should be noted that, in the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0043] It will be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0044] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0045] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1The function specified in one or more boxes.

[0046] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 The steps for the function specified in one or more boxes.

[0047] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0048] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A method for identifying mileage fraud at turnaround points in online ride-hailing vehicles based on a multi-dimensional model, characterized by: The specific steps include: S101. Collect driving data and trajectory data from online ride-hailing services and perform error checking and correction on the driving data. Add a sliding window to extract the average eigenvalue of the driving data. Use a decision tree to build a driving behavior model to classify actual driving behavior. S102. Establish a driver credit score system and add or subtract driver credit scores based on passenger reviews, number of completed orders, and passenger complaints; S103. Collect data on the cities where ride-hailing services are located, divide the city data into regular 5-square-meter grids, count the number of points that meet the requirements, use Euclidean methods to determine the distance between the location point and the nearest road network, and construct a city point quality model using a decision tree to classify weather, tunnels, and mountains. S104. Based on the classification of actual driving behavior in the driving behavior model, the expected behavior of the trajectory data is obtained. The expected quality score of each location point in the expected behavior of the trajectory data is evaluated in combination with the city point quality model. The quality scores of the actual driving trajectories are compared, and the difference is obtained as an outlier for judgment. The credibility value of the actual driving trajectory is used as an influencing factor to recalculate the driver's credit score.

2. The multi-dimensional model-based method for identifying mileage fraud at turnaround points in online ride-hailing services according to claim 1, characterized in that: In S101, the driving data and trajectory data of the online-hailing car are collected, and the error values ​​in the driving data of the online-hailing car are checked. For the driving speed and driving direction, it is checked whether they exceed the reasonable range. If they exceed the reasonable range, the data is deleted. For the distance between adjacent driving points and the GPS signal strength during the driving process, it is checked whether there are errors and inconsistent information. A sliding window is added to apply a window with a fixed size of 5 minutes to the driving data, and the window slides with a fixed step size of half the window length. The granularity of the driving data is controlled by adjusting the window size and step size. The sliding window is used to extract the average characteristic values ​​of the distance between adjacent driving points, driving speed, driving direction and GPS signal strength during driving. The decision tree is used to construct a driving behavior model to classify the actual driving behavior.

3. The method for identifying mileage fraud at turnaround points in online ride-hailing services based on a multi-dimensional model according to claim 2, characterized in that: The specific steps of constructing the driving behavior model using a decision tree are as follows: selecting the average characteristic value of the distance between adjacent driving points, driving speed, driving direction, and GPS signal strength during driving as the root node of the decision tree, and dividing the driving data into different driving subsets. For each non-leaf node, the best feature among the average characteristic values ​​of the distance between adjacent driving points, driving speed, driving direction, and GPS signal strength during driving is selected for splitting. When the Gini index of the non-leaf node is smaller, the impurity of the node is judged to be lower, and the driving data category represented by the feature is more consistent, the current node is split into multiple child nodes based on the selected best feature, wherein each child node corresponds to a different driving subset under the feature value. The selection and splitting steps are repeated until the driving data represented by the node corresponding to the feature belongs to the same category and is labeled with the category, including normal driving, speeding, sudden braking, and abnormal turning.

4. The multi-dimensional model-based method for identifying mileage fraud at turnaround points in online ride-hailing services according to claim 1, characterized in that: In S102, a driver credit score system is established. The default value of the credit score for a newly registered driver is 50 points. With respect to the passenger evaluation dimension, points are added or subtracted according to the star rating of the passenger evaluation. If a passenger gives a 5-star rating, the driver's credit score is added by 0.5 points; if a passenger gives a 4-star rating, the driver's credit score is added by 0.2 points; if a passenger gives a 3-star rating, the driver's credit score is added by 0 points; if a passenger gives a 2-star rating, the driver's credit score is subtracted by 0.3 points; and if a passenger gives a 1-star rating, the driver's credit score is subtracted by 0.5 points. With respect to the number of completed orders, points are calculated based on the number of orders the driver completes without any fault. The driver's credit score is added by 3 points for every 100 orders completed. With respect to the passenger complaint dimension, if a customer complains and it is confirmed that the driver is responsible, the driver's credit score is subtracted by 2 points. If the cumulative number of customer complaints and it is confirmed that the driver is responsible is between 3 and 5, the driver's credit score is subtracted by 3 points each time. If the cumulative number of customer complaints and it is confirmed that the driver is responsible is between 3 and 5, the driver's credit score is subtracted by 4 points each time. If the cumulative number of customer complaints and it is confirmed that the driver is responsible exceeds 10 times, the driver's credit score is subtracted by 5 points each time.

5. The method for identifying mileage fraud at turnaround points in online ride-hailing services based on a multi-dimensional model according to claim 1, characterized in that: In S103, the city data of the online car-hailing service is collected, including weather, tunnels, and mountains. Combined with the GPS signal strength during the driving process, the satellite map is connected to check the erroneous values ​​in the city data and replace them. The city data is divided into regular grid areas of 5 square meters. In each grid, the number of points that meet the conditions is counted according to the weather, tunnels, and mountains. For each location point, the spatial index structure is used to accelerate the nearest search, and the Euclidean method is used to obtain the distance between the location point and the nearest road network. The distance between the location point and the nearest road network and the number of points are used as feature values, and a decision tree is used to construct a city point quality model to classify weather, tunnels, and mountains.

6. The multi-dimensional model-based method for identifying mileage fraud at turnaround points in online ride-hailing services according to claim 5, characterized in that: The specific Euclidean formula is: Among them, (X1, Y1) represents the coordinates of the location point in the grid area, (X2, Y2) represents the coordinates of the nearest road network in the grid area, and S represents the distance between the location point and the nearest road network.

7. The method for identifying mileage fraud at turnaround points in online ride-hailing services based on a multi-dimensional model according to claim 5, characterized in that: The specific steps of constructing the urban point quality model using a decision tree are as follows: selecting the characteristic values ​​represented by the distance between the location point and the nearest road network and the number of points as the root node of the decision tree, and dividing the urban data into different urban subsets. For each non-leaf node, the best feature among the distance between the location point and the nearest road network and the number of points is selected for splitting. When the Gini index of the non-leaf node is smaller, the impurity of the node is judged to be lower, and the urban data category represented by the feature is more consistent, the current node is split into multiple child nodes according to the selected best feature, where each child node corresponds to a different urban subset under the feature value. The selection and splitting steps are repeated until the urban data represented by the node corresponding feature belongs to the same category and is labeled with the category, including weather, tunnel and mountain.

8. The method for identifying mileage fraud at turnaround points in online ride-hailing services based on a multi-dimensional model according to claim 1, characterized in that: In said S104, the expected behavior of the trajectory data is obtained according to the classification of the actual driving behavior of the driving behavior model, and the expected quality score of each position point in the expected behavior of the trajectory data is evaluated in combination with the urban point quality model. The expected quality score is compared with the quality score of the actual driving trajectory, and the difference is obtained as the abnormal value, which ranges from [0-10]. The abnormal value range of [0-3] is considered normal, the abnormal value range of [3-5] is considered slightly abnormal, the abnormal value range of [5-8] is considered abnormal, and the abnormal value range of [8-10] is considered seriously abnormal. The credibility value of the actual driving trajectory is evaluated according to the driver's credit score, and the credibility value of the actual driving trajectory is used as an influencing factor to recalculate the driver's credit score. The credibility value of the machine credit score is [0-10]. The credibility value range of the driving trajectory is [0-2], which is considered untrustworthy. The credibility value range of the driving trajectory is [2-5], which is considered slightly untrustworthy. The credibility value range of the driving trajectory is [5-8], which is considered slightly trustworthy. The credibility value range of the driving trajectory is [8-10], which is considered trustworthy. When the credibility value range of the driving trajectory is [0-2], the driver's credibility score is reduced by 2 points. When the credibility value range of the driving trajectory is [2-5], the driver's credibility score is reduced by 1 point. The driving trajectory in the untrustworthy range is marked for settlement and submitted to the manual review team. The review results are used as the influencing factors of the driver's credibility score system and the driver's score is adjusted.

9. The multi-dimensional model-based method for identifying mileage fraud at turnaround points in online ride-hailing services according to claim 8, characterized in that: The specific formula for recalculating the driver's credit score using the actual driving trajectory credibility value as an influencing factor is: Among them, P represents the credibility value of the actual driving trajectory, t represents the abnormal value, e represents the driver's credit score value, and r represents the impact factor.

Images