Method and device for identifying abnormal user of online car-hailing

Abstract

The invention belongs to the technical field of online car-hailing, and discloses an online car-hailing abnormal user identification method and device. The method comprises the steps of obtaining a current incoming parameter of a current request interface corresponding to a current request of a current session corresponding to a current order and a previous incoming parameter of a previous request interface corresponding to a previous request in an online car-hailing order placing process of a user; the input parameters comprise a vehicle using stroke, a vehicle using place and vehicle using time; inputting the current input parameter and the previous input parameter into a pre-constructed order monitoring model, and outputting a target similarity of a current interface of the current session and whether the target similarity is lower than an interval left end point of a similarity qualified interval; and if the target similarity is lower than the interval left end point of the similarity qualified interval, accumulating the number of exceptions, and when the number of exceptions is greater than a preset threshold value of the number of exceptions, determining that the user is an abnormal user. According to the invention, the recognition accuracy of the abnormal user can be improved.

Description

Technical Field

[0001] The present application relates to the field of online car-hailing technology, and in particular to a method and device for identifying abnormal users of online car-hailing. Background Art

[0002] In online car-hailing platforms, when users are ordering cars, the server will receive requests from normal users and feedback the car-hailing information to the users based on the requests. However, in addition to receiving requests from normal users, it will also receive requests from a large number of abnormal users such as robots, which fabricate order parameters and swipe a large amount of important data from the interface. In the prior art, abnormal users are detected by user login frequency and IP address, and the detection method is relatively simple, and the accuracy of abnormal user identification is low. Therefore, there is an urgent need for a method for identifying abnormal users of online car-hailing to improve the accuracy of abnormal user identification. Summary of the invention

[0003] Based on this, it is necessary to provide a method and device for identifying abnormal users of online ride-hailing services in response to the above technical problems.

[0004] In a first aspect, a method for identifying abnormal users of online car-hailing services is provided, the method comprising: When a user places an online car-hailing order, obtain the current input parameter of the current request interface corresponding to the current request of the current session corresponding to the current order and the previous input parameter of the previous request interface corresponding to the previous request; the input parameters include the car trip, car location and car time; Input the current input parameter and the previous input parameter into a pre-built order monitoring model, and output the target similarity of the current interface of the current session, and whether the target similarity is lower than the left endpoint of the similarity qualified interval; If the target similarity is lower than the left end point of the similarity qualified interval, the number of abnormalities is accumulated, and when the number of abnormalities is greater than a preset threshold of the number of abnormalities, the user is determined to be an abnormal user.

[0005] As an optional implementation manner, the inputting the current input parameter and the previous input parameter into a pre-built order monitoring model and outputting the target similarity of the current interface of the current session includes: In a pre-built order monitoring model, a ratio of the previous input parameter to the current input parameter is determined as an initial similarity of the current interface of the current session; The product value of the initial similarity and the predetermined similarity correction coefficient is determined as the target similarity of the current interface of the current session.

[0006] As an optional implementation, the method further includes: Obtain order log data for multiple orders within a first preset time period; The mean of the last input parameter in each order in the order log data is determined, and the ratio of a preset constant threshold and the mean is determined as a correction coefficient of the similarity.

[0007] As an optional implementation, the method further includes: Obtain a historical order sequence within a second preset time period, and an input parameter of each request interface of each of the historical orders in the historical order sequence; For any request interface of any historical order, the ratio of the input parameter of the previous request interface of the request interface to the input parameter of the request interface is determined as the initial similarity of the request interface of the historical order, and the product value of the initial similarity and a predetermined similarity correction coefficient is determined as the target similarity of the request interface; When determining the similarity qualified interval of the target request interface of the target historical order of the historical order sequence, the mean and variance of the target similarity are determined according to the target similarities of each request interface corresponding to the first historical order to the target historical order, and the difference between the mean and the variance is determined as the left endpoint of the similarity qualified interval, and the sum of the mean and the variance is determined as the right endpoint of the similarity qualified interval; the first historical order is the first historical order of the historical order sequence.

[0008] As an optional implementation, the method further includes: Obtain a training sample set, the training sample set including multiple training samples and training interval results corresponding to each of the training samples, the training sample is an input parameter of any request interface of multiple training history orders, and the training interval result is a similarity qualified interval; Based on the training samples and the training interval results corresponding to the training samples, the initial order monitoring model is trained to obtain the trained order monitoring model.

[0009] As an optional implementation, the method further includes: Obtaining a test sample set, the test sample set including multiple test samples and test sample results corresponding to each of the test samples, the test sample being an input parameter of any request interface of multiple test history orders, and the test sample result being a similarity qualified interval; Input each of the inspection samples into the trained initial order monitoring model, and output the inspection similarity interval corresponding to each inspection sample; Determining the detection accuracy of the trained initial order monitoring model according to the similarity qualified interval and the test similarity interval; If the detection accuracy is greater than the accuracy threshold, it is determined that the trained initial order monitoring model meets the preset training requirements; otherwise, it is determined that the trained initial order monitoring model does not meet the preset training requirements.

[0010] As an optional implementation, the method further includes: If the target similarity is within the similarity qualified interval, it indicates that the user is a normal user and the current request is a normal request; If the target similarity is higher than the right end point of the similarity qualified interval, it indicates that the user is a normal user and the current request is a repeated request.

[0011] In a second aspect, a device for identifying abnormal users of online car-hailing services is provided, the device comprising: The first acquisition module is used to obtain the current input parameter of the current request interface corresponding to the current request of the current session and the previous input parameter of the previous request interface corresponding to the previous request during the process of the user placing an online car-hailing order; the input parameters include the car trip, car location and car time; An input module, used for inputting the current input parameter and the previous input parameter into a pre-built order monitoring model, and outputting the target similarity of the current interface of the current session, and whether the target similarity is lower than the left endpoint of the similarity qualified interval; The accumulation module is used to accumulate the number of abnormalities if the target similarity is lower than the left end point of the similarity qualified interval, and when the number of abnormalities is greater than a preset abnormal number threshold, determine that the user is an abnormal user.

[0012] As an optional implementation manner, the input module is specifically used for: In a pre-built order monitoring model, a ratio of the previous input parameter to the current input parameter is determined as an initial similarity of the current interface of the current session; The product value of the initial similarity and the predetermined similarity correction coefficient is determined as the target similarity of the current interface of the current session.

[0013] As an optional implementation, the device further includes: A second acquisition module is used to acquire order log data of multiple orders within a first preset time period; The first determination module is used to determine the mean of the last input parameter in each order in the order log data, and determine the ratio of a preset constant threshold and the mean as a correction coefficient of the similarity.

[0014] As an optional implementation, the device further includes: A third acquisition module is used to acquire a historical order sequence within a second preset time period, and an input parameter of each request interface of each historical order in the historical order sequence; The second determination module is used to determine, for any request interface of any historical order, a ratio of an input parameter of a previous request interface of the request interface to an input parameter of the request interface as an initial similarity of the request interface of the historical order, and determine a product value of the initial similarity and a predetermined similarity correction coefficient as a target similarity of the request interface; The third determination module is used to determine the similarity qualified interval of the target request interface of the target historical order of the historical order sequence, based on the target similarities of each request interface corresponding to the first historical order to the target historical order, determine the mean and variance of the target similarity, and determine the difference between the mean and the variance as the left endpoint of the similarity qualified interval, and determine the sum of the mean and the variance as the right endpoint of the similarity qualified interval; the first historical order is the first historical order of the historical order sequence.

[0015] As an optional implementation, the device further includes: A fourth acquisition module is used to acquire a training sample set, wherein the training sample set includes multiple training samples and training interval results corresponding to each of the training samples, wherein the training sample is an input parameter of any request interface of multiple training history orders, and the training interval result is a similarity qualified interval; The training module is used to train the initial order monitoring model based on each of the training samples and the training interval results corresponding to each of the training samples to obtain the trained order monitoring model.

[0016] As an optional implementation, the device further includes: A fifth acquisition module is used to acquire a test sample set, wherein the test sample set includes multiple test samples and test sample results corresponding to each of the test samples, wherein the test sample is an input parameter of any request interface of multiple test history orders, and the test sample result is a similarity qualified interval; An output module, used to input each of the inspection samples into the trained initial order monitoring model, and output the inspection similarity interval corresponding to each inspection sample; A fourth determination module, used to determine the detection accuracy of the trained initial order monitoring model according to the similarity qualified interval and the test similarity interval; A determination module is used to determine whether the trained initial order monitoring model meets the preset training requirements if the detection accuracy is greater than the accuracy threshold, otherwise determine whether the trained initial order monitoring model does not meet the preset training requirements.

[0017] As an optional implementation, the device further includes: A first characterization module, configured to characterize the user as a normal user and the current request as a normal request if the target similarity is within the similarity qualified interval; The second characterization module is used to characterize that the user is a normal user and the current request is a repeated request if the target similarity is higher than the right end point of the similarity qualified interval.

[0018] In a third aspect, a system for identifying abnormal users of an online car-hailing service is provided, and the system for identifying abnormal users of an online car-hailing service comprises: a method for identifying abnormal users of an online car-hailing service as described in the first aspect and a device for identifying abnormal users of an online car-hailing service as described in the second aspect.

[0019] The present application provides a method for identifying abnormal users of online ride-hailing services. The technical solution provided by the embodiments of the present application brings at least the following beneficial effects: in the process of a user placing an order for an online ride-hailing service, the current input parameter of the current request interface corresponding to the previous request and the previous input parameter of the previous request interface corresponding to the previous request are obtained. According to the order monitoring model, the current input parameter and the previous input parameter, the target similarity of the current interface is determined, and it is determined whether the target similarity is within the qualified similarity interval. When the number of times the target similarity is lower than the left endpoint of the qualified similarity interval exceeds the preset abnormal number threshold, the user is determined to be an abnormal user. By analyzing the implicit association patterns between the user behavior sequence, the input content, and the context environment, a more accurate abnormality determination is achieved. The method for identifying abnormal users is enriched, and the accuracy of identifying abnormal users is improved.

[0020] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0022] Figure 1 A schematic diagram of the structure of a system for identifying abnormal users of online ride-hailing vehicles provided in an embodiment of the present application; Figure 2 A flowchart of a method for identifying abnormal users of online ride-hailing services provided in an embodiment of the present application; Figure 3 A schematic diagram of the structure of a device for identifying abnormal users of an online ride-hailing service provided in an embodiment of the present application. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0024] The method for identifying abnormal users of online car-hailing provided in the embodiment of the present application can be applied to the system for identifying abnormal users of online car-hailing. Figure 1 As shown, the identification system for abnormal users of online car-hailing includes a controller 101 and a user terminal 102. The controller 101 and the user terminal 102 are connected.

[0025] The controller 101 is used to obtain the current input parameters of the current request interface corresponding to the current request of the current session and the previous input parameters of the previous request interface corresponding to the previous request sent by the user terminal 102 during the process of the user placing an online car-hailing order; the input parameters include the car trip, car location and car time. The current input parameters and the previous input parameters are input into the pre-built order monitoring model, and the target similarity of the current interface of the current session is output, and whether the target similarity is lower than the left endpoint of the similarity qualified interval. If the target similarity is lower than the left endpoint of the similarity qualified interval, the number of abnormalities is accumulated. When the number of abnormalities is greater than the preset abnormality threshold, the user is determined to be an abnormal user.

[0026] The user terminal 102 is used to send the current input parameters of the current request interface corresponding to the current request of the current session corresponding to the current order to the controller 101 during the process of placing an online car-hailing order.

[0027] The following will describe in detail a method for identifying abnormal users of online car-hailing services provided by an embodiment of the present application in conjunction with a specific implementation method. Figure 2 A flowchart of a method for identifying abnormal users of online car-hailing provided in an embodiment of the present application, such as Figure 2 As shown, the specific steps are as follows: Step 201, when the user places an online car-hailing order, obtain the current input parameters of the current request interface corresponding to the current request of the current session corresponding to the current order and the previous input parameters of the previous request interface corresponding to the previous request; the input parameters include the car trip, car location and car time.

[0028] In practice, in the process of ordering cars, in addition to receiving requests from normal users, the server of online car-hailing services often receives a large number of requests from abnormal users such as robots, such as constantly fabricating order parameters to swipe a large number of important data from the interface. Important data can be map data, price data, etc. Therefore, it is necessary to identify abnormal users of online car-hailing services and identify abnormal users. After identifying abnormal users, the session requests of abnormal users can be automatically rejected to avoid leakage of important information. As the analysis of the ordering process of users in booking car services shows that in the order of the same online car-hailing service, the content before and after the car-hailing information of normal users is related and has great similarity. The car-hailing information includes the car-hailing itinerary, the car-hailing location and the car-hailing time. When the same user uses a car, the car-hailing itinerary, the car-hailing location and the car-hailing time are usually similar. Therefore, by analyzing the implicit association pattern between the user behavior sequence, the input content and the context environment, it is possible to determine whether the user is an abnormal user, and achieve a more accurate abnormal user determination. Therefore, when a user places an online car-hailing order, the current input parameter of the current request interface corresponding to the current request of the current session and the previous input parameter of the previous request interface corresponding to the previous request can be obtained. The input parameters include the car trip, car location and car time. The subsequent step can determine the similarity of the two input parameters to determine whether the user is an abnormal user.

[0029] Step 202, input the current input parameter and the previous input parameter into the pre-built order monitoring model, output the target similarity of the current interface of the current session, and whether the target similarity is lower than the left endpoint of the similarity qualified interval.

[0030] In implementation, after obtaining the current input parameter of the current request interface corresponding to the current request and the previous input parameter of the previous request interface corresponding to the previous request, it is necessary to determine the similarity between the current input parameter and the previous input parameter to determine whether the user is an abnormal user. The current input parameter and the previous input parameter can be input into a pre-built order monitoring model, and the order monitoring model outputs the target similarity of the current interface of the current session, and whether the target similarity is lower than the left endpoint of the similarity qualified interval. The similarity qualified interval is the interval in which the similarity of the input parameters of the two requests before and after a normal user places an order for an online car-hailing service is in. If the target similarity is within the similarity qualified interval, the user is determined to be a normal user.

[0031] Furthermore, the specific process of inputting the current input parameter and the previous input parameter into the pre-built order monitoring model and outputting the target similarity of the current interface of the current session is as follows: Step 1: In the pre-built order monitoring model, the ratio of the previous input parameter to the current input parameter is determined as the initial similarity of the current interface of the current session.

[0032] In the implementation, after the current input parameter and the previous input parameter are input into the pre-built order monitoring model, the ratio of the previous input parameter to the current input parameter is determined as the initial similarity of the current interface of the current session in the pre-built order monitoring model. The input parameter of the interface refers to the data that needs to be passed in when calling the interface, and the output parameter refers to the data returned to the caller after calling the interface. Usually, the input parameter and the output parameter are defined in a certain data format, such as JSON or XML. The input parameter is a string, and the server can directly calculate the initial similarity of the two strings. When calling the interface, the data needs to be passed into the interface according to the input parameter format of the interface. The interface will perform corresponding processing based on the input data and return data that conforms to the output parameter format to the caller. The definition of input and output parameters is very important in interface design, because different interfaces may require different data types and data structures.

[0033] Step 2: determine the product value of the initial similarity and the predetermined similarity correction coefficient as the target similarity of the current interface of the current session.

[0034] In the implementation, when the user places an online car-hailing order, the amount of information in the car-hailing information input will affect the similarity of the matching of the two input parameters. Since many requests will be initiated during the entire ordering process, each request corresponds to an interface request, and it is not necessary to complete the entire process to complete the order. In the same order, the more information the request interface has, the higher the similarity of the request interface, so it is necessary to correct the position of the previous and next requests of the order. If not corrected, the range of similarity obtained in the next step will be different when the request position is different, and the corresponding order monitoring model needs to be calculated for each request position, and a qualified similarity interval cannot be determined for different request positions. Therefore, correction is required. After the correction, the range of similarity obtained in the next step is the same when the request position is at any position, and only one order monitoring model needs to be determined. For example, when the request position n=1, 2, 3, the number of parameters is 11, 12, 13 respectively, the amount of one parameter is 10, and the amount of data is 110, 120, 130 respectively. The larger n is, the greater the similarity. If no correction is made first, for example, among multiple orders, the mean similarity of multiple orders with n=1 is 50%, and the similarity range is 45%~55%. The mean similarity of multiple orders with n=3 is 70%, and the similarity range is 65%~75%. That is to say, when the request position n is different, the mean and variance of the similarity corresponding to each n must be calculated. Therefore, correction is required, and the mean and variance of the similarity are the same when n is any value. The product value of the initial similarity and the predetermined similarity correction coefficient can be determined as the target similarity of the current interface of the current session.

[0035] Furthermore, the steps for determining the correction coefficient of similarity are as follows: Step three, obtaining order log data for multiple orders within a first preset time period.

[0036] In practice, the purpose of the correction coefficient is to make the range of similarity of the request interface at each position the same. Then, the similarity of the request interface at each request position in the order can be replaced with the range of similarity of the request interface at the last request position. Therefore, when determining the correction coefficient of similarity, the order log data of multiple orders within the first preset time length is obtained. Subsequently, the input parameters of the last request interface of each order are obtained from the order log data.

[0037] Step 4: determine the mean of the last input parameter in each order in the order log data, and determine the ratio of the preset constant threshold and the mean as the correction coefficient of the similarity.

[0038] In implementation, the last input parameter of each order is obtained from the order log data, and then the mean of the last input parameter of each order in the order log data is determined, and the ratio of the preset constant threshold and the mean is determined as the correction coefficient of the similarity. In practical applications, the preset constant threshold can be any constant, and for the convenience of calculation, the preset constant threshold can be 1. In this way, the similarity of each position request in the order can be divided into a unified standard measurement interval.

[0039] Step 203: if the target similarity is lower than the left end point of the similarity qualified interval, the number of abnormalities is accumulated. When the number of abnormalities is greater than a preset threshold of the number of abnormalities, the user is determined to be an abnormal user.

[0040] In implementation, if the target similarity is lower than the left endpoint of the similarity qualified interval, it means that the similarity between the current input parameter of the current request and the previous input parameter of the previous request of the user is low, and the user may be an abnormal user. In order to avoid contingency, the number of abnormalities can be accumulated. If the number of abnormalities is large, the user is determined to be an abnormal user. Therefore, if the target similarity is lower than the left endpoint of the similarity qualified interval, the number of abnormalities is accumulated. The number of abnormalities is compared with the preset abnormal number threshold. When the number of abnormalities is greater than the preset abnormal number threshold, the user is determined to be an abnormal user. In this way, the determination of abnormal users will not be too arbitrary, reducing contingency.

[0041] Furthermore, in addition to the left endpoint of the interval where the target similarity is lower than the similarity qualified interval, there are two other cases where the target similarity is within the similarity qualified interval and the right endpoint of the interval where the target similarity is higher than the similarity qualified interval: Method 1: If the target similarity is within the qualified similarity range, it indicates that the user is a normal user and the current request is a normal request.

[0042] In implementation, the target similarity is compared with the left end point and the right end point of the similarity qualified interval respectively. If the target similarity is within the similarity qualified interval, it indicates that the user is a normal user and the current request is a normal request.

[0043] Method 2: If the target similarity is higher than the right end point of the similarity qualified interval, it indicates that the user is a normal user and the current request is a repeated request.

[0044] In implementation, the target similarity is compared with the left endpoint and the right endpoint of the similarity qualified interval respectively. If the target similarity is higher than the right endpoint of the similarity qualified interval, it indicates that the user is a normal user and the current request is a repeated request.

[0045] Furthermore, the steps for determining the similarity qualified interval are as follows: Step A, obtaining a historical order sequence within a second preset time period, and input parameters of each request interface of each historical order in the historical order sequence.

[0046] In implementation, when determining the similarity qualified interval of similarity, the historical order sequence within the second preset time length and the input parameters of each request interface of each historical order in the historical order sequence can be obtained. The subsequent step determines the similarity of each request interface of each historical order, and then determines the similarity qualified interval.

[0047] Step B, for any request interface of any historical order, the ratio of the input parameters of the previous request interface of the request interface to the input parameters of the request interface is determined as the initial similarity of the request interface of the historical order, and the product value of the initial similarity and the predetermined similarity correction coefficient is determined as the target similarity of the request interface.

[0048] In implementation, after obtaining the input parameters of each request interface of each historical order in the historical order sequence, it is necessary to determine the similarity of each request interface of each historical order. That is, for any request interface of any historical order, the ratio of the input parameters of the previous request interface of the request interface and the input parameters of the request interface is determined as the initial similarity of the request interface of the historical order, and the product value of the initial similarity and the pre-determined similarity correction coefficient is determined as the target similarity of the request interface. The subsequent steps will determine the similarity qualified interval based on each target similarity.

[0049] Step C, when determining the similarity qualified interval of the target request interface of the target historical order of the historical order sequence, determine the mean and variance of the target similarity according to the target similarities of each request interface corresponding to the first historical order to the target historical order, and determine the difference between the mean and the variance as the left endpoint of the similarity qualified interval, and determine the sum of the mean and the variance as the right endpoint of the similarity qualified interval; the first historical order is the first historical order of the historical order sequence.

[0050] In implementation, when determining the similarity qualified interval of the target request interface of the target historical order of the historical order sequence, the mean and variance of the target similarities of all request interfaces of all historical orders are determined based on the target similarities of each request interface corresponding to the first historical order to the target historical order. The difference between the mean and the variance is determined as the left endpoint of the similarity qualified interval, and the sum of the mean and the variance is determined as the right endpoint of the similarity qualified interval. In this way, the composed similarity qualified interval can include the similarity of the previous and next parameters when a normal user places an online car-hailing order. Among them, the first historical order is the first historical order in the historical order sequence, and the target historical order can be the last historical order in the historical order sequence.

[0051] Furthermore, once the similarity qualified interval is determined, it is unlikely to change. However, if the interface of the order process changes due to demand and the input parameters change, the similarity qualified interval may change.

[0052] Furthermore, the training process of the order monitoring model is as follows: Step D, obtaining a training sample set, the training sample set includes multiple training samples and training interval results corresponding to each training sample, the training sample is an input parameter of any request interface of multiple training history orders, and the training interval result is a similarity qualified interval.

[0053] In the implementation, a training sample set is obtained, which includes multiple training samples and training interval results corresponding to each training sample. The training sample is an input parameter of any request interface of multiple training historical orders, and the training interval result is a similarity qualified interval. The subsequent step is to train the initial order monitoring model through the training sample set.

[0054] Step E: Based on each training sample and the training interval result corresponding to each training sample, the initial order monitoring model is trained to obtain a trained order monitoring model.

[0055] In the implementation, each training sample is input into the initial order monitoring model, the sample similarity interval corresponding to each training sample is output, and the sample similarity interval is compared with the qualified similarity interval. If the sample similarity interval and the qualified similarity interval are the same, it means that the initial order monitoring model recognizes correctly. If they are not the same, it means that the recognition is wrong. According to the sample similarity interval and the qualified similarity interval of each training sample, the recognition accuracy of the initial order monitoring model is determined. The recognition accuracy is compared with the accuracy threshold. If the recognition accuracy is greater than the accuracy threshold, it means that the training of the initial order monitoring model is completed. If the recognition accuracy is less than or equal to the accuracy threshold, the model parameters of the initial order monitoring model are adjusted, and the steps of inputting each training sample into the adjusted initial order monitoring model and outputting the sample similarity interval corresponding to each training sample are repeated until the recognition accuracy is greater than the accuracy threshold, indicating that the adjusted order monitoring model training is completed, and a trained order monitoring model is obtained.

[0056] Furthermore, after the model training is completed, it needs to be tested. The specific test steps are as follows: Step a, obtaining a test sample set, the test sample set includes multiple test samples and test sample results corresponding to each test sample, the test sample is an input parameter of any request interface of multiple test history orders, and the test sample result is a similarity qualified interval.

[0057] In the implementation, a test sample set is obtained, which includes multiple test samples and test sample results corresponding to each test sample. The test sample is an input parameter of any request interface of multiple test history orders, and the test sample result is a similarity qualified interval. The subsequent steps can test the trained initial order monitoring model according to the test sample set.

[0058] Step b: input each inspection sample into the trained initial order monitoring model, and output the inspection similarity interval corresponding to each inspection sample.

[0059] In the implementation, each test sample is input into the trained initial order monitoring model, and the test similarity interval corresponding to each test sample is output. Subsequently, it can be determined whether the initial order monitoring model meets the training requirements based on the test similarity interval and the similarity qualified interval.

[0060] Step c: determining the detection accuracy of the trained initial order monitoring model according to the qualified similarity interval and the tested similarity interval.

[0061] In implementation, if the similarity qualified interval and the test similarity interval are the same, it is determined that the detection of the trained initial order monitoring model is accurate. If the similarity qualified interval and the test similarity interval are not the same, it is determined that the detection of the trained initial order monitoring model is inaccurate. Therefore, the detection accuracy of the trained initial order monitoring model can be determined based on the similarity qualified interval and the test similarity interval.

[0062] Step d: if the detection accuracy is greater than the accuracy threshold, it is determined that the trained initial order monitoring model meets the preset training requirements; otherwise, it is determined that the trained initial order monitoring model does not meet the preset training requirements.

[0063] In implementation, the detection accuracy is greater than the accuracy threshold, and if the detection accuracy is greater than the accuracy threshold, it is determined that the trained initial order monitoring model meets the preset training requirements. If the detection accuracy is less than or equal to the accuracy threshold, it is determined that the trained initial order monitoring model does not meet the preset training requirements.

[0064] An embodiment of the present application provides a method for identifying abnormal users of online ride-hailing services. During the process of a user placing an online ride-hailing order, the current input parameter of the current request interface corresponding to the previous request and the previous input parameter of the previous request interface corresponding to the previous request are obtained. According to the order monitoring model, the current input parameter and the previous input parameter, the target similarity of the current interface is determined, and it is determined whether the target similarity is within the qualified similarity interval. When the number of times the target similarity is lower than the left endpoint of the qualified similarity interval exceeds the preset abnormal number threshold, the user is determined to be an abnormal user. By analyzing the implicit association patterns between the user behavior sequence, the input content, and the context environment, a more accurate abnormality determination can be achieved. The method for identifying abnormal users is enriched, and the accuracy of identifying abnormal users is improved.

[0065] It should be understood that although Figure 2 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 2 At least part of the steps may include multiple steps or multiple stages. These steps or stages are not necessarily performed at the same time, but can be performed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed in turn or alternately with other steps or at least part of the steps or stages in other steps.

[0066] It can be understood that the same / similar parts between the various embodiments of the above method in this specification can refer to each other, and each embodiment focuses on the differences from other embodiments. For related points, please refer to the description of other method embodiments.

[0067] The present application also provides a device for identifying abnormal users of online car-hailing services. Figure 3 As shown, the device comprises: The first acquisition module 301 is used to obtain the current input parameter of the current request interface corresponding to the current request of the current session and the previous input parameter of the previous request interface corresponding to the previous request during the process of the user placing an online car-hailing order; the input parameters include the car trip, car location and car time; An input module 302 is used to input the current input parameter and the previous input parameter into a pre-built order monitoring model, and output the target similarity of the current interface of the current session, and whether the target similarity is lower than the left endpoint of the similarity qualified interval; The accumulation module 303 is used to accumulate the number of abnormalities if the target similarity is lower than the left end point of the similarity qualified interval, and when the number of abnormalities is greater than a preset abnormal number threshold, determine that the user is an abnormal user.

[0068] As an optional implementation manner, the input module 302 is specifically used for: In a pre-built order monitoring model, a ratio of the previous input parameter to the current input parameter is determined as an initial similarity of the current interface of the current session; The product value of the initial similarity and the predetermined similarity correction coefficient is determined as the target similarity of the current interface of the current session.

[0069] As an optional implementation, the device further includes: A second acquisition module is used to acquire order log data of multiple orders within a first preset time period; The first determination module is used to determine the mean of the last input parameter in each order in the order log data, and determine the ratio of a preset constant threshold and the mean as a correction coefficient of the similarity.

[0070] As an optional implementation, the device further includes: A third acquisition module is used to acquire a historical order sequence within a second preset time period, and an input parameter of each request interface of each historical order in the historical order sequence; The second determination module is used to determine, for any request interface of any historical order, a ratio of an input parameter of a previous request interface of the request interface to an input parameter of the request interface as an initial similarity of the request interface of the historical order, and determine a product value of the initial similarity and a predetermined similarity correction coefficient as a target similarity of the request interface; The third determination module is used to determine the similarity qualified interval of the target request interface of the target historical order of the historical order sequence, based on the target similarities of each request interface corresponding to the first historical order to the target historical order, determine the mean and variance of the target similarity, and determine the difference between the mean and the variance as the left endpoint of the similarity qualified interval, and determine the sum of the mean and the variance as the right endpoint of the similarity qualified interval; the first historical order is the first historical order of the historical order sequence.

[0071] As an optional implementation, the device further includes: A fourth acquisition module is used to acquire a training sample set, wherein the training sample set includes multiple training samples and training interval results corresponding to each of the training samples, wherein the training sample is an input parameter of any request interface of multiple training history orders, and the training interval result is a similarity qualified interval; The training module is used to train the initial order monitoring model based on each of the training samples and the training interval results corresponding to each of the training samples to obtain the trained order monitoring model.

[0072] As an optional implementation, the device further includes: A fifth acquisition module is used to acquire a test sample set, wherein the test sample set includes multiple test samples and test sample results corresponding to each of the test samples, wherein the test sample is an input parameter of any request interface of multiple test history orders, and the test sample result is a similarity qualified interval; An output module, used for inputting each of the inspection samples into the trained initial order monitoring model, outputting the inspection similarity interval corresponding to each inspection sample, and whether the inspection similarity is lower than the left endpoint of the interval of the similarity qualified interval; A fourth determination module, used to determine the detection accuracy of the trained initial order monitoring model according to the similarity qualified interval and the test similarity interval; A determination module is used to determine whether the trained initial order monitoring model meets the preset training requirements if the detection accuracy is greater than the accuracy threshold, otherwise determine whether the trained initial order monitoring model does not meet the preset training requirements.

[0073] As an optional implementation, the device further includes: A first characterization module, configured to characterize the user as a normal user and the current request as a normal request if the target similarity is within the similarity qualified interval; The second characterization module is used to characterize that the user is a normal user and the current request is a repeated request if the target similarity is higher than the right end point of the similarity qualified interval.

[0074] An embodiment of the present application provides a device for identifying abnormal users of online ride-hailing. During the process of a user placing an online ride-hailing order, the current input parameter of the current request interface corresponding to the previous request and the previous input parameter of the previous request interface corresponding to the previous request are obtained. According to the order monitoring model, the current input parameter and the previous input parameter, the target similarity of the current interface is determined, and it is determined whether the target similarity is within the qualified similarity interval. When the number of times the target similarity is lower than the left endpoint of the qualified similarity interval exceeds the preset abnormal number threshold, the user is determined to be an abnormal user. By analyzing the implicit association patterns between the user behavior sequence, the input content, and the context environment, a more accurate abnormality determination is achieved. The method for identifying abnormal users is enriched, and the accuracy of identifying abnormal users is improved.

[0075] For the specific limitations of the device for identifying abnormal users of online ride-hailing vehicles, please refer to the limitations of the method for identifying abnormal users of online ride-hailing vehicles mentioned above, which will not be repeated here. Each module in the above-mentioned device for identifying abnormal users of online ride-hailing vehicles can be implemented in whole or in part through software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0076] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0077] It should also be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for display, data for analysis, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0078] Each embodiment in this specification is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0079] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0080] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A method for identifying abnormal users of online car-hailing, characterized in that: The method comprises: When a user places an online car-hailing order, obtain the current input parameter of the current request interface corresponding to the current request of the current session corresponding to the current order and the previous input parameter of the previous request interface corresponding to the previous request; the input parameters include the car trip, car location and car time; Input the current input parameter and the previous input parameter into a pre-built order monitoring model, and output the target similarity of the current interface of the current session, and whether the target similarity is lower than the left endpoint of the similarity qualified interval; If the target similarity is lower than the left end point of the similarity qualified interval, the number of abnormalities is accumulated, and when the number of abnormalities is greater than a preset threshold of the number of abnormalities, the user is determined to be an abnormal user.

2. The method according to claim 1, characterized in that The step of inputting the current input parameter and the previous input parameter into a pre-built order monitoring model and outputting the target similarity of the current interface of the current session includes: In a pre-built order monitoring model, a ratio of the previous input parameter to the current input parameter is determined as an initial similarity of the current interface of the current session; The product value of the initial similarity and the predetermined similarity correction coefficient is determined as the target similarity of the current interface of the current session.

3. The method according to claim 2, characterized in that The method further comprises: Obtain order log data for multiple orders within a first preset time period; The mean of the last input parameter in each order in the order log data is determined, and the ratio of a preset constant threshold and the mean is determined as a correction coefficient of the similarity.

4. The method according to claim 1, characterized in that: The method further comprises: Obtain a historical order sequence within a second preset time period, and input parameters of each request interface of each of the historical orders in the historical order sequence; For any request interface of any historical order, the ratio of the input parameter of the previous request interface of the request interface to the input parameter of the request interface is determined as the initial similarity of the request interface of the historical order, and the product value of the initial similarity and a predetermined similarity correction coefficient is determined as the target similarity of the request interface; When determining the similarity qualified interval of the target request interface of the target historical order of the historical order sequence, the mean and variance of the target similarity are determined according to the target similarities of each request interface corresponding to the first historical order to the target historical order, and the difference between the mean and the variance is determined as the left endpoint of the similarity qualified interval, and the sum of the mean and the variance is determined as the right endpoint of the similarity qualified interval; the first historical order is the first historical order of the historical order sequence.

5. The method according to claim 1, characterized in that The method further comprises: Obtain a training sample set, the training sample set including multiple training samples and training interval results corresponding to each of the training samples, the training sample is an input parameter of any request interface of multiple training history orders, and the training interval result is a similarity qualified interval; Based on the training samples and the training interval results corresponding to the training samples, the initial order monitoring model is trained to obtain the trained order monitoring model.

6. The method according to claim 5, characterized in that The method further comprises: Obtaining a test sample set, the test sample set including multiple test samples and test sample results corresponding to each of the test samples, the test sample being an input parameter of any request interface of multiple test history orders, and the test sample result being a similarity qualified interval; Input each of the inspection samples into the trained initial order monitoring model, and output the inspection similarity interval corresponding to each inspection sample; Determining the detection accuracy of the trained initial order monitoring model according to the similarity qualified interval and the test similarity interval; If the detection accuracy is greater than the accuracy threshold, it is determined that the trained initial order monitoring model meets the preset training requirements; otherwise, it is determined that the trained initial order monitoring model does not meet the preset training requirements.

7. The method according to claim 1, characterized in that The method further comprises: If the target similarity is within the similarity qualified interval, it indicates that the user is a normal user and the current request is a normal request; If the target similarity is higher than the right end point of the similarity qualified interval, it indicates that the user is a normal user and the current request is a repeated request.

8. A device for identifying abnormal users of online car-hailing, characterized in that: The device comprises: The first acquisition module is used to obtain the current input parameter of the current request interface corresponding to the current request of the current session and the previous input parameter of the previous request interface corresponding to the previous request during the process of the user placing an online car-hailing order; the input parameters include the car trip, car location and car time; An input module, used for inputting the current input parameter and the previous input parameter into a pre-built order monitoring model, and outputting the target similarity of the current interface of the current session, and whether the target similarity is lower than the left endpoint of the interval of the similarity qualified interval; The accumulation module is used to accumulate the number of abnormalities if the target similarity is lower than the left end point of the similarity qualified interval, and when the number of abnormalities is greater than a preset abnormal number threshold, determine that the user is an abnormal user.

9. The device according to claim 8, characterized in that The input module is specifically used for: In a pre-built order monitoring model, a ratio of the previous input parameter to the current input parameter is determined as an initial similarity of the current interface of the current session; The product value of the initial similarity and the predetermined similarity correction coefficient is determined as the target similarity of the current interface of the current session.

10. The device according to claim 8, characterized in that The device also includes: A second acquisition module is used to acquire order log data of multiple orders within a first preset time period; The first determination module is used to determine the mean of the last input parameter in each order in the order log data, and determine the ratio of a preset constant threshold and the mean as a correction coefficient of the similarity.

Images