Responsibility judgment method and system, electronic equipment and storage medium

Abstract

The invention provides a responsibility judgment method and system, electronic equipment and a storage medium, and the method comprises the steps: obtaining an order, and determining a driver related to the order and the driver information; obtaining a pre-constructed driver white list, and performing preliminary responsibility judgment on the driver according to the driver white list and the driver information to obtain a preliminary judgment result; wherein the driver white list is constructed according to a pre-constructed driver responsibility probability prediction model and historical order data; if the preliminary judgment result indicates that the driver has responsibility, obtaining order information of the order, and determining algorithm label information of each sub-model according to the order information by using each pre-constructed sub-model; wherein the algorithm label information is null or comprises at least one algorithm label; and determining a plurality of hit strategies according to each algorithm label and each other algorithm label, and performing final responsibility judgment on the driver according to each hit strategy to obtain a final judgment result.

Description

Technical Field

[0001] This invention relates to the field of transportation technology, and more specifically, to a method, system, electronic device, and storage medium for determining liability. Background Technology

[0002] With the continuous development of society, vehicles are being used more and more widely, especially in the transportation industry. During transportation, in order to ensure user experience and transportation safety, liability can be determined for orders with abnormalities.

[0003] In existing technologies, orders with abnormalities are generally reviewed manually to determine the corresponding responsibility. That is, the driver is held responsible for the abnormality of the order. However, manual review is not only inaccurate, but also requires a lot of manpower and is inefficient. Summary of the Invention

[0004] In view of this, the present invention provides a method, system, electronic device and storage medium for determining liability, with the aim of improving the accuracy of liability determination for orders, avoiding the expenditure of a large amount of manpower and improving the efficiency of liability determination.

[0005] The first aspect of this application provides a method for determining liability, the method comprising:

[0006] Obtain the order and identify the driver and their information associated with the order;

[0007] Obtain a pre-built driver whitelist, and make a preliminary determination of the driver's responsibility based on the driver whitelist and the driver information to obtain a preliminary determination result; wherein, the driver whitelist is constructed based on a pre-built driver responsibility probability prediction model and historical order data;

[0008] If the preliminary determination result indicates that the driver is at fault, the order information of the order is obtained, and the algorithm tag information of each sub-model is determined based on the order information using each pre-built sub-model; wherein, the algorithm tag information is empty, or includes at least one algorithm tag;

[0009] Multiple hit strategies are determined based on each of the aforementioned algorithm tags and other algorithm tags, and the driver's final responsibility is determined based on each of the aforementioned hit strategies to obtain the final determination result.

[0010] Optionally, constructing the driver whitelist based on a pre-built driver liability probability prediction model and historical order data includes:

[0011] Obtain historical order data, wherein the historical order data includes multiple historical orders;

[0012] For each historical order, obtain historical order information related to the historical order, wherein the historical order information includes historical driver information of historical drivers, historical user information of historical users, and the actual probability of the historical driver being at fault.

[0013] The historical driver information and the historical user information are input into a pre-built driver liability probability prediction model, which then makes a prediction based on the historical driver information and the historical user information to obtain the historical driver's liability probability. The driver liability probability prediction model is trained on a driver liability probability prediction model to be trained based on the historical driver information, historical user information, and the actual liability probability of the historical driver related to each historical order.

[0014] A driver whitelist is constructed based on the driver IDs of historical drivers whose probability of being at fault is greater than a preset probability.

[0015] Optionally, the step of obtaining a pre-built driver whitelist, and making a preliminary determination of the driver's responsibility based on the driver whitelist and the driver information to obtain a preliminary determination result includes:

[0016] Obtain a pre-built driver whitelist, wherein the driver whitelist includes multiple driver IDs;

[0017] Determine whether the driver ID in the driver information is in the driver whitelist;

[0018] If the location is specified, a preliminary determination result indicating that the driver is at fault is generated;

[0019] If not located, a preliminary determination result indicating that the driver is not at fault is generated.

[0020] Optionally, the order information of the order is obtained, and the algorithm label information of each sub-model is determined based on the order information using each pre-built sub-model, including:

[0021] Obtain the order information of the order, wherein the order information includes at least the privacy number recording order remarks;

[0022] For each pre-built sub-model, the privacy number recording order notes are input into the sub-model, and the sub-model analyzes the privacy number recording order notes and outputs corresponding behavior label information. The behavior label information can be empty or at least one behavior label. The sub-model is trained using an algorithm that matches historical privacy number recording notes with the scene corresponding to the sub-model.

[0023] If the behavior label information is empty, generate algorithm label information indicating that it is empty;

[0024] If the behavior label information is not empty, based on the pre-set association between behavior labels and algorithm labels, determine the algorithm label that matches each behavior label in the behavior label information from the pre-set algorithm labels, and generate the algorithm label information of the sub-model based on the algorithm label that matches each behavior label.

[0025] Optionally, the step of determining multiple target hitting strategies based on each of the algorithm tags, and making a final responsibility determination on the driver based on each of the hitting strategies to obtain a final determination result, includes:

[0026] Obtain multiple pre-set other tags, and combine each of the algorithm tags and each of the other tags to obtain multiple algorithm combinations;

[0027] For each of the algorithm combinations, a target hitting strategy that matches the algorithm combination is selected from a set of pre-defined hitting strategies;

[0028] The highest priority target hitting strategy is selected from all the target hitting strategies, and the judgment result corresponding to the highest priority target hitting strategy is used as the final judgment result for the driver.

[0029] Optionally, the method further includes:

[0030] Determine whether the final determination indicates that the driver is at fault;

[0031] If the final determination indicates that the driver is at fault, determine the corresponding appeal label for the driver;

[0032] If the appeal tag is a full appeal, output the corresponding appeal steps, and detect in real time whether the driver has filed an appeal for the order;

[0033] If the appeal tag is not a full appeal, output the corresponding appeal pop-up window, and detect in real time whether there is an appeal operation by the driver for the order;

[0034] If such an order exists, it will be sent to a human review panel.

[0035] A second aspect of this application provides a liability determination system, the system comprising:

[0036] The acquisition module is used to acquire orders and identify the drivers and their information associated with the orders.

[0037] The preliminary liability determination module is used to obtain a pre-built driver whitelist, and to make a preliminary liability determination on the driver based on the driver whitelist and the driver information to obtain a preliminary determination result; wherein, the driver whitelist is constructed by the driver whitelist construction module based on a pre-built driver liability probability prediction model and historical order data;

[0038] An algorithm tag determination module is used to, if the preliminary judgment result indicates that the driver is responsible, obtain the order information of the order, and determine the algorithm tag information of each of the pre-built sub-models based on the order information; wherein, the algorithm tag information is empty, or includes at least one algorithm tag;

[0039] The final responsibility determination module is used to determine multiple hit strategies based on each of the algorithm tags and other algorithm tags, and to make a final responsibility determination on the driver based on each of the hit strategies, so as to obtain the final determination result.

[0040] Optionally, the driver whitelist construction module is specifically used for:

[0041] Obtain historical order data, wherein the historical order data includes multiple historical orders;

[0042] For each historical order, obtain historical order information related to the historical order, wherein the historical order information includes historical driver information of historical drivers, historical user information of historical users, and the actual probability of the historical driver being at fault.

[0043] The historical driver information and the historical user information are input into a pre-built driver liability probability prediction model, which then makes a prediction based on the historical driver information and the historical user information to obtain the historical driver's liability probability. The driver liability probability prediction model is trained on a driver liability probability prediction model to be trained based on the historical driver information, historical user information, and the actual liability probability of the historical driver related to each historical order.

[0044] A driver whitelist is constructed based on the driver IDs of historical drivers whose probability of being at fault is greater than a preset probability.

[0045] A third aspect of this application provides an electronic device, comprising: a processor and a memory, the processor and the memory being connected via a bus; wherein the processor is configured to call and execute a program stored in the memory; and the memory is configured to store the program, the program being configured to implement the liability determination method provided in the first aspect of this application.

[0046] The fourth aspect of this application provides a computer-readable storage medium storing computer-executable instructions for performing the liability determination method provided in the first aspect of this application.

[0047] This application provides a liability determination method, system, electronic device, and storage medium. The method involves: acquiring an order and identifying the driver and their information associated with the order; acquiring a pre-built driver whitelist; performing a preliminary liability determination on the driver based on the whitelist and driver information to obtain a preliminary determination result; wherein the driver whitelist is constructed based on a pre-built driver liability probability prediction model and historical order data; if the preliminary determination result indicates driver liability, acquiring the order information and determining the algorithm label information for each pre-built sub-model based on the order information; wherein the algorithm label information is either empty or includes at least one algorithm label; determining multiple hit strategies based on each algorithm label and other algorithm labels, and performing a final liability determination on the driver based on each hit strategy to obtain a final determination result. The technical means provided in this application pre-constructs a driver whitelist using a pre-built driver liability probability prediction model and historical order data. This whitelist is then used to make a preliminary liability determination on the driver based on order-related driver information. Only when the preliminary liability determination indicates that the driver is liable is the final liability determination made. This effectively reduces the possibility of a large number of non-liability orders being manually reviewed, avoiding the expenditure of a large amount of manpower and thus improving the efficiency of liability determination. Multiple sub-models are pre-constructed. When it is determined that the preliminary liability determination indicates that the driver is liable, each sub-model is used to determine the matching algorithm label based on the order information. This allows the use of each algorithm label and other algorithm labels to determine the hitting strategy. Finally, the final liability determination on the driver is completed based on the determined hitting strategy. This not only improves machine coverage but also improves the accuracy of liability determination. Attached Figure Description

[0048] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0049] Figure 1 A flowchart illustrating a liability determination method provided in an embodiment of this application;

[0050] Figure 2 An example diagram illustrating a preliminary liability determination provided in an embodiment of this application;

[0051] Figure 3 An example diagram illustrating a final liability determination provided in an embodiment of this application;

[0052] Figure 4 This is a schematic diagram of the structure of a liability determination system provided in an embodiment of this application;

[0053] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0054] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0055] In this application, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0056] See Figure 1 The diagram illustrates a flowchart of a liability determination method provided in this application, which specifically includes the following steps:

[0057] S101: Obtain the order and identify the driver and his / her information associated with the order.

[0058] During the specific execution of step S101, it is possible to detect in real time whether there are any abnormal orders; if so, the order can be obtained and analyzed to identify the driver who received the order, so as to further obtain the driver's information.

[0059] It should be noted that the driver's information may include at least the driver's ID.

[0060] In some embodiments, orders that are received in less than a preset time can be identified as abnormal orders. For example, if an order is canceled in less than a minute after being received, it can be identified as an abnormal order.

[0061] The above is only a preferred method for determining whether an order is abnormal in this application. The specific method for determining whether an order is abnormal can be set according to the actual application, and is not limited in this embodiment of the application.

[0062] S102: Obtain the pre-built driver whitelist, make a preliminary determination of the driver's responsibility based on the driver whitelist and driver information, and obtain a preliminary determination result; wherein, the driver whitelist is constructed based on the pre-built driver liability probability prediction model and historical order data.

[0063] In this embodiment of the application, in order to reduce the amount of manual review, the driver liability probability prediction model to be trained can be trained in advance using historical order data to obtain the driver liability probability prediction model. In order to use the driver liability probability prediction model to predict the liability probability of the historical driver responsible for each historical order, and to construct a corresponding driver whitelist based on each historical driver and its liability probability, wherein the driver whitelist includes multiple driver IDs; after obtaining the driver information, the driver can be preliminarily responsible for the driver based on the driver ID in the driver information and each driver ID in the driver whitelist, and a preliminary judgment result can be obtained, wherein the preliminary judgment result indicates whether the driver is responsible or not responsible.

[0064] Optionally, the process of constructing a driver whitelist based on a pre-built driver liability probability prediction model and historical order data can be as follows: Obtain historical order data, which includes multiple historical orders; for each historical order, obtain historical order information related to that order, including historical driver information, historical user information, and the actual liability probability of the historical driver; input the historical driver information and historical user information into the pre-built driver liability probability prediction model, enabling the model to predict the liability probability of the historical driver based on the historical driver information and historical user information; wherein the driver liability probability prediction model is trained using the historical driver information, historical user information, and the actual liability probability of the historical driver related to each historical order; and construct a corresponding driver whitelist based on the driver IDs of historical drivers whose liability probability is greater than a preset probability.

[0065] It should be noted that the preset probability can be 20%, 25%, or 30%, and the corresponding preset probability can be set according to the actual application. This application embodiment does not limit it.

[0066] For example, assuming a preset probability of 20%, after predicting the probability of each historical driver being at fault, the driver ID of at least one historical driver with a probability of being at fault of less than 20% can be added to a pre-built driver whitelist.

[0067] Optionally, the process of obtaining a pre-built driver whitelist and making a preliminary determination of driver liability based on the driver whitelist and driver information can be as follows: obtain a pre-built driver whitelist, which includes multiple driver IDs; determine whether the driver ID in the driver information is in the driver whitelist; if it is, generate a preliminary determination result indicating that the driver is liable; if it is not, generate a preliminary determination result indicating that the driver is not liable.

[0068] In some embodiments, historical driver information, historical user information, and the actual liability probability of historical drivers related to each historical order can be pre-input into the driver liability probability prediction model to be trained. The driver liability probability prediction model to be trained then makes predictions based on the historical driver information and historical user information related to each historical order to obtain the liability probability of the historical driver related to that historical order. The parameters of the driver liability probability prediction model to be trained are adjusted with the goal of making the liability probability of the historical driver approach the actual liability probability of the historical driver. The parameters of the driver liability probability prediction model to be trained are then adjusted until the parameters of the driver liability probability prediction model to be trained converge, thus obtaining the corresponding driver liability probability prediction model.

[0069] In practical applications, see Figure 2 The system can pre-acquire historical order data; for each historical order in the historical order data, it can acquire related historical order information; extract multiple historical driver features from the historical driver information in the historical order information, including the driver's personal information, activity level, and performance; extract multiple historical user features from the historical user information in the historical order information, including the user's personal information, order completion status, and customer complaint status; input the extracted historical driver features and historical user features into a pre-trained driver liability probability prediction model, enabling the model to predict the driver's liability probability based on the input historical driver and user features; based on the liability probabilities of each historical driver, at least one historical driver that meets the pre-set rules is selected from the historical drivers, and a corresponding driver whitelist is constructed based on the driver IDs of the selected historical drivers, wherein the pre-set rules indicate the selection of historical drivers with a liability probability greater than a preset probability.

[0070] After obtaining the driver ID of the driver associated with the current order, it can be determined whether there is a driver ID with the same driver ID in the pre-built driver registration list. If it exists, it can be considered that the driver is not responsible for the order abnormality, and a preliminary judgment result indicating that the driver is not responsible can be generated. If it does not exist, it can be considered that the driver may be responsible for the order abnormality, and a preliminary judgment result indicating that the driver is responsible can be generated.

[0071] After obtaining the initial judgment result of the driver, the order can be stored as a historical order. After the final judgment result of the order is determined, the actual probability of the driver's liability and the whitelist effect evaluation (precision recall rate, missed judgment rate and false judgment rate) can be calculated based on the final judgment result. In order to optimize the driver liability probability prediction model by using the driver information, user information and the actual probability of the driver's liability related to the order, so as to ensure and improve the prediction accuracy of the driver liability probability prediction model.

[0072] S103: Determine whether the preliminary judgment result indicates that the driver is at fault; if the preliminary judgment result indicates that the driver is at fault, proceed to step S104.

[0073] In the specific execution of step S103, after determining the preliminary judgment result of the driver using the pre-built driver whitelist, it can be further determined whether the preliminary judgment result indicates that the driver is responsible; if the preliminary judgment result indicates that the driver is responsible, step S104 is executed; if the preliminary judgment result does not indicate that the driver is responsible, that is, it indicates that the driver is not responsible, the preliminary judgment result can be used as the final judgment result of the driver, and the final judgment result of the driver can be output to avoid the order flowing into manual review.

[0074] Therefore, it can be seen that by building a driver whitelist in advance, a portion of the orders for which the driver is not at fault can be effectively filtered out. This not only reduces the number of orders that need to be manually reviewed and improves the efficiency of liability determination, but also solves the problem that current technology cannot directly output the driver's judgment results and can only rely on manual review, thus increasing the coverage of the machine.

[0075] S104: Obtain the order information of the order and use the pre-built sub-model to determine the algorithm label information of each sub-model based on the order information.

[0076] In the specific execution of step S104, if the driver is initially determined to be at fault, in order to further improve the accuracy of the fault determination, the order information can be further obtained, including at least the privacy number recording order notes.

[0077] For each pre-built sub-model, the privacy number recording order notes are input into the sub-model, enabling the sub-model to analyze the privacy number recording order notes and output corresponding behavior label information. The behavior label information can be empty, or at least one behavior label. The sub-model is trained using an algorithm that matches historical privacy number recording notes with the corresponding scene of the sub-model. If the behavior label information is empty, an algorithm label indicating emptiness is generated. If the behavior label information is not empty, based on the pre-set association between behavior labels and algorithm labels, the algorithm label matching each behavior label in the behavior label information is determined from the pre-set algorithm labels, and the algorithm label information of the sub-model is generated based on the algorithm label matching each behavior label.

[0078] It should be noted that the privacy number recording order notes include the privacy number recording of the order and the order notes.

[0079] In this embodiment, to improve the accuracy of liability determination and enhance the coverage of machine liability determination, while maintaining the machine's responsible recall rate, a corresponding strategy set can be pre-collected to analyze multiple responsibility scenarios and decompose multiple algorithm tags from each responsibility scenario. Multiple applicable algorithms are determined based on each algorithm tag, and a behavior tag corresponding to each algorithm tag is set, along with the association between the algorithm tag and its behavior tag. Historical privacy number recordings between historical drivers and historical users related to each historical order are obtained. For each algorithm, the algorithm is used to analyze each historical privacy number recording, obtaining at least one behavior tag for each historical order corresponding to that historical privacy number recording. A corresponding loss function is constructed using one behavior tag and at least one actual behavior tag of the historical order. The constructed loss function is used to adjust the algorithm's parameters until the algorithm converges, resulting in a corresponding sub-model, i.e., a sub-model corresponding to the responsibility scenario of the algorithm.

[0080] After obtaining the order information, for each sub-model, the privacy number recording order remarks from the obtained order information are input into the sub-model, so that the sub-model can analyze the input privacy number recording order remarks and output the corresponding behavior tag information. If the behavior tag information is empty, an algorithm tag information indicating emptiness is generated. If the behavior tag information is not empty, the algorithm tag that is associated with each behavior tag in the behavior tag information is determined from the pre-set algorithm tags, thereby determining at least one algorithm tag corresponding to the sub-model.

[0081] It should be noted that the multiple liability scenarios analyzed from the strategy set may include liability scenarios such as "driver's personal reasons", "wrong order", "driver raising price", and "skipping the order"; the behavior tags may be wrong order, raising price, skipping the order, etc., which are not limited in this embodiment of the application.

[0082] It should also be noted that the pre-built sub-models can be dialect models, personal cause models, wrong-robbery models, etc., and corresponding sub-models can be built according to actual applications. This application embodiment does not limit them.

[0083] In summary, by simultaneously inputting the privacy number recording and order remarks of an order into multiple sub-models for different responsibility scenarios, the sub-models output their respective extracted behavior labels based on the input privacy number recording and order remarks, and determine the corresponding algorithm label for each behavior label, thereby determining the corresponding algorithm label for the sub-model.

[0084] S105: Determine multiple hit strategies based on each algorithm label and other algorithm labels, and make a final responsibility determination for the driver based on each hit strategy to obtain the final determination result.

[0085] In the embodiments of this application, multiple hit strategies are preset, wherein each hit strategy includes at least one hit algorithm tag and / or at least one non-hit algorithm tag; the priority and corresponding judgment result of each hit strategy can also be preset.

[0086] For example, the pre-set hit strategy can be set to hit algorithm tag 1, algorithm tag 2, and algorithm tag 3, but not algorithm tag 4; or it can be set to hit algorithm tag 1, or tag 3, or algorithm tag 4 (other algorithm tags); or it can be set to hit algorithm tag 1, algorithm tag 2, and algorithm tag 3. The corresponding hit strategy can be set according to actual needs, and this application embodiment does not limit it.

[0087] In the specific execution step S105, after determining the algorithm label information corresponding to each sub-model, empty algorithm label information is removed, and each algorithm label in the non-empty algorithm label information is determined; multiple other algorithm labels are obtained, and each algorithm label and each other label are combined to obtain multiple algorithm combinations; for each algorithm combination, a target hitting strategy matching the algorithm combination is selected from the pre-set hitting strategies; the target hitting strategy with the highest priority is selected from the target hitting strategies, and the judgment result corresponding to the target hitting strategy with the highest priority is used as the driver's final judgment result.

[0088] It should be noted that at least one other algorithm tag can be randomly extracted from the pre-set algorithm tags, in addition to the algorithm tags in the non-empty algorithm tag information. Alternatively, all other algorithm tags besides the algorithm tags in the non-empty algorithm tag information can be obtained. This application embodiment does not limit this.

[0089] In practical applications, see Figure 3 Suppose that the privacy number recording order notes are input into pre-set sub-models, where each sub-model includes at least dialect, personal reasons, and mistakenly taken orders. Each sub-model analyzes the input privacy number recording order notes and outputs corresponding behavior tag information. If the behavior tag information is empty, an empty algorithm tag is generated. If the behavior tag information is not empty, based on the pre-set association between behavior tags and algorithm tags, the algorithm tag matching each behavior tag in the behavior tag information is determined from the pre-set algorithm tags, so as to determine at least one algorithm tag corresponding to the sub-model.

[0090] Other tags are acquired and each algorithm tag is combined with other tags to determine the target hit tag for each algorithm combination from the pre-set hit tags. The highest effective priority strategy (the highest priority hit strategy) is selected from the hit tags matched by each algorithm combination, and the judgment result of the highest effective priority strategy is used as the final judgment result of the driver (liable / not liable).

[0091] In summary, to further improve the accuracy of liability determination, when the preliminary determination indicates that the driver is at fault, the pre-built sub-models are used to extract corresponding algorithm tags based on the order's privacy number, audio recording, and order remarks. The final determination of the driver's liability is then determined based on the algorithm tags and pre-set hit strategies. This not only improves machine coverage but also directly outputs the corresponding determination results without relying on manual review, thereby reducing the number of orders requiring manual review and improving the efficiency of liability determination.

[0092] Furthermore, in this embodiment of the application, after obtaining the final judgment result of the order, it can be further determined whether the final judgment result indicates that the driver is responsible; if the final judgment result indicates that the driver is responsible, the driver's corresponding appeal tag is determined; if the appeal tag is a full appeal, the corresponding appeal steps are output, and it is detected in real time whether there is a driver making an appeal operation for the order; if the appeal tag is not a full appeal, the corresponding appeal pop-up is output, and it is detected in real time whether there is a driver making an appeal operation for the order; if so, the order is sent to manual review.

[0093] It should be noted that, in order to further improve the efficiency of order liability determination, this application may further add a corresponding appeal scenario determination mechanism so that drivers can appeal the order for which liability has been determined when they receive a liability determination notification.

[0094] In some embodiments, because some drivers may not be familiar with the correct appeal procedures, leading to errors in the appeal process, in order to improve the efficiency of driver appeals, when the final judgment of the order indicates that the driver is at fault, the driver's corresponding appeal tag can be further determined. If the driver's appeal tag is "full appeal," it means that the driver is not familiar with the correct appeal procedures and often makes mistakes in the appeal process. In this case, the corresponding appeal steps can be further output so that the driver can complete the corresponding appeal based on the suggested steps. At the same time, it can also remind the driver to learn the relevant steps before appealing to avoid errors in the appeal process. If it is not "full appeal," the driver can be considered a driver with poor appeal skills. The test can directly output the corresponding appeal pop-up so that the driver can directly appeal based on the appeal pop-up.

[0095] This application provides a method for determining liability, which involves obtaining an order and identifying the driver and their information associated with the order; obtaining a pre-built driver whitelist; performing a preliminary liability determination on the driver based on the driver whitelist and driver information to obtain a preliminary determination result; wherein the driver whitelist is constructed based on a pre-built driver liability probability prediction model and historical order data; if the preliminary determination result indicates that the driver is liable, obtaining the order information and determining the algorithm label information of each pre-built sub-model based on the order information; wherein the algorithm label information is empty or includes at least one algorithm label; determining multiple hit strategies based on each algorithm label and other algorithm labels, and performing a final liability determination on the driver based on each hit strategy to obtain a final determination result. The technical means provided in this application pre-constructs a driver whitelist using a pre-built driver liability probability prediction model and historical order data. This whitelist is then used to make a preliminary liability determination on the driver based on order-related driver information. Only when the preliminary liability determination indicates that the driver is liable is the final liability determination made. This effectively reduces the possibility of a large number of non-liability orders being manually reviewed, avoiding the expenditure of a large amount of manpower and thus improving the efficiency of liability determination. Multiple sub-models are pre-constructed. When it is determined that the preliminary liability determination indicates that the driver is liable, each sub-model is used to determine the matching algorithm label based on the order information. This allows the use of each algorithm label and other algorithm labels to determine the hitting strategy. Finally, the final liability determination on the driver is completed based on the determined hitting strategy. This not only improves machine coverage but also improves the accuracy of liability determination.

[0096] Based on the liability determination method shown in the above embodiments of the present invention, correspondingly, the present invention also shows a liability determination system, such as... Figure 4 As shown, the system includes:

[0097] The acquisition module 41 is used to acquire orders and identify the drivers and their information associated with the orders.

[0098] The preliminary liability determination module 42 is used to obtain a pre-built driver whitelist, and to make a preliminary liability determination on the driver based on the driver whitelist and driver information to obtain a preliminary determination result; wherein, the driver whitelist is constructed by the driver whitelist construction module based on a pre-built driver liability probability prediction model and historical order data;

[0099] The algorithm label determination module 43 is used to obtain the order information of the order if the preliminary judgment result indicates that the driver is responsible, and to determine the algorithm label information of each sub-model based on the order information using each pre-built sub-model; wherein, the algorithm label information is empty, or includes at least one algorithm label;

[0100] The final responsibility determination module 44 is used to determine multiple hit strategies based on each algorithm label and other algorithm labels, and to make a final responsibility determination for the driver based on each hit strategy, so as to obtain the final determination result.

[0101] This application provides a liability determination system, which involves obtaining an order and identifying the driver and their information associated with the order; obtaining a pre-built driver whitelist; performing a preliminary liability determination on the driver based on the driver whitelist and driver information to obtain a preliminary determination result; wherein the driver whitelist is constructed based on a pre-built driver liability probability prediction model and historical order data; if the preliminary determination result indicates that the driver is liable, obtaining the order information and using each pre-built sub-model to determine the algorithm label information of each sub-model based on the order information; wherein the algorithm label information is empty or includes at least one algorithm label; determining multiple hit strategies based on each algorithm label and other algorithm labels, and performing a final liability determination on the driver based on each hit strategy to obtain a final determination result. The technical means provided in this application pre-constructs a driver whitelist using a pre-built driver liability probability prediction model and historical order data. This whitelist is then used to make a preliminary liability determination on the driver based on order-related driver information. Only when the preliminary liability determination indicates that the driver is liable is the final liability determination made. This effectively reduces the possibility of a large number of non-liability orders being manually reviewed, avoiding the expenditure of a large amount of manpower and thus improving the efficiency of liability determination. Multiple sub-models are pre-constructed. When it is determined that the preliminary liability determination indicates that the driver is liable, each sub-model is used to determine the matching algorithm label based on the order information. This allows the use of each algorithm label and other algorithm labels to determine the hitting strategy. Finally, the final liability determination on the driver is completed based on the determined hitting strategy. This not only improves machine coverage but also improves the accuracy of liability determination.

[0102] Optional, a driver whitelist building module, specifically used for:

[0103] Retrieve historical order data, which includes multiple historical orders;

[0104] For each historical order, obtain the historical order information related to the historical order, including the historical driver information of the historical driver, the historical user information of the historical user, and the actual probability of the historical driver being at fault.

[0105] Historical driver information and historical user information are input into a pre-built driver liability probability prediction model, which then makes predictions based on the historical driver information and historical users to obtain the historical driver liability probability. The driver liability probability prediction model is trained on the driver liability probability prediction model to be trained based on the historical driver information, historical user information and the actual liability probability of the historical driver related to each historical order.

[0106] A driver whitelist is constructed based on the driver IDs of historical drivers whose probability of being at fault is greater than a preset probability.

[0107] Optional, preliminary responsibility determination module, specifically used for:

[0108] Retrieve a pre-built driver whitelist, which includes multiple driver IDs;

[0109] Determine whether the driver ID in the driver information is on the driver whitelist;

[0110] If the location is specified, a preliminary determination result indicating that the driver is at fault will be generated;

[0111] If not located, a preliminary determination result indicating that the driver is not at fault is generated.

[0112] Optional, the algorithm label determination module is specifically used for:

[0113] Obtain the order information, which includes at least the privacy number recording order notes;

[0114] For each pre-built sub-model, the privacy number recording order notes are input into the sub-model, so that the sub-model can analyze the privacy number recording order notes and output the corresponding behavior label information. The behavior label information can be empty or at least one behavior label. The sub-model is trained by an algorithm that uses historical privacy number recording notes and the scene matching algorithm corresponding to the sub-model.

[0115] If the behavior label information is empty, generate algorithm label information indicating that it is empty;

[0116] If the behavior label information is not empty, based on the pre-set association between behavior labels and algorithm labels, determine the algorithm label that matches each behavior label in the behavior label information from the pre-set algorithm labels, and generate the algorithm label information of the sub-model based on the algorithm label that matches each behavior label.

[0117] Optional, the final liability determination module is specifically used for:

[0118] Obtain multiple pre-set other tags, and combine each algorithm tag with each other tag to obtain multiple algorithm combinations;

[0119] For each algorithm combination, select the target hitting strategy that matches the algorithm combination from the pre-set hitting strategies;

[0120] The highest priority target hitting strategy is selected from all target hitting strategies, and the judgment result corresponding to the highest priority target hitting strategy is used as the driver's final judgment result.

[0121] Optionally, the liability determination system provided in this application also includes an appeal module;

[0122] The appeals module is used to determine whether the final judgment indicates that the driver is at fault;

[0123] If the final determination indicates that the driver is at fault, determine the driver's corresponding appeal label;

[0124] If the appeal tag is "full appeal", output the corresponding appeal steps and detect in real time whether there are any drivers making appeals for the order;

[0125] If the appeal tag is not "full appeal", output the corresponding appeal pop-up window and check in real time whether there are any appeal operations by drivers regarding the order;

[0126] If such an order exists, it will be sent to a human review process.

[0127] This application also provides a storage medium storing program instructions, which, when loaded and executed by a processor, implement any of the above-described embodiments of the liability determination method.

[0128] This application also provides an electronic device, such as Figure 5 As shown, the device includes a processor 501 and a memory 502, which are connected via a bus; the memory stores program instructions; the processor calls the program instructions in the memory to execute any of the above-described embodiments of the liability determination method.

[0129] The processor mentioned in this article can be the terminal's CPU, an integrated MCU within the terminal, or a combination of a CPU and an MCU. Furthermore, the processor contains a kernel that retrieves the corresponding program from memory; one or more kernels can be configured.

[0130] The memory may include non-permanent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0131] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, for system or system embodiments, since they are basically similar to method embodiments, the description is relatively simple, and relevant parts can be referred to the descriptions in the method embodiments. The systems and system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0132] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0133] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0134] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for determining liability, characterized in that, The method includes: Obtain the order and identify the driver and their information associated with the order; Obtain a pre-built driver whitelist, and make a preliminary determination of the driver's responsibility based on the driver whitelist and the driver information to obtain a preliminary determination result; wherein, the driver whitelist is constructed based on a pre-built driver responsibility probability prediction model and historical order data; If the preliminary determination result indicates that the driver is at fault, the order information of the order is obtained, and the algorithm tag information of each sub-model is determined based on the order information using each pre-built sub-model; wherein, the algorithm tag information is empty, or includes at least one algorithm tag; Multiple hit strategies are determined based on each of the aforementioned algorithm tags and other algorithm tags, and the driver's final responsibility is determined based on each of the aforementioned hit strategies to obtain the final determination result.

2. The method according to claim 1, characterized in that, The process of constructing the driver whitelist based on a pre-built driver liability probability prediction model and historical order data includes: Obtain historical order data, wherein the historical order data includes multiple historical orders; For each historical order, obtain historical order information related to the historical order, wherein the historical order information includes historical driver information of historical drivers, historical user information of historical users, and the actual probability of the historical driver being at fault. The historical driver information and the historical user information are input into a pre-built driver liability probability prediction model, which then makes a prediction based on the historical driver information and the historical user information to obtain the historical driver's liability probability. The driver liability probability prediction model is trained on a driver liability probability prediction model to be trained based on the historical driver information, historical user information, and the actual liability probability of the historical driver related to each historical order. A driver whitelist is constructed based on the driver IDs of historical drivers whose probability of being at fault is greater than a preset probability.

3. The method according to claim 1, characterized in that, The process of obtaining a pre-built driver whitelist, and making a preliminary liability determination on the driver based on the driver whitelist and the driver information to obtain a preliminary determination result includes: Obtain a pre-built driver whitelist, wherein the driver whitelist includes multiple driver IDs; Determine whether the driver ID in the driver information is in the driver whitelist; If the location is specified, a preliminary determination result indicating that the driver is at fault is generated; If not located, a preliminary determination result indicating that the driver is not at fault is generated.

4. The method according to claim 1, characterized in that, Obtain the order information of the order, and determine the algorithm label information of each sub-model based on the order information using each pre-built sub-model, including: Obtain the order information of the order, wherein the order information includes at least the privacy number recording order remarks; For each pre-built sub-model, the privacy number recording order notes are input into the sub-model, and the sub-model analyzes the privacy number recording order notes and outputs corresponding behavior label information. The behavior label information can be empty or at least one behavior label. The sub-model is trained using an algorithm that matches historical privacy number recording notes with the scene corresponding to the sub-model. If the behavior label information is empty, generate algorithm label information indicating that it is empty; If the behavior label information is not empty, based on the pre-set association between behavior labels and algorithm labels, determine the algorithm label that matches each behavior label in the behavior label information from the pre-set algorithm labels, and generate the algorithm label information of the sub-model based on the algorithm label that matches each behavior label.

5. The method according to claim 1, characterized in that, The process of determining multiple target-hitting strategies based on each of the algorithm tags, and then making a final responsibility determination for the driver based on each of the hit strategies to obtain a final determination result includes: Obtain multiple pre-set other tags, and combine each of the algorithm tags and each of the other tags to obtain multiple algorithm combinations; For each of the algorithm combinations, a target hitting strategy that matches the algorithm combination is selected from a set of pre-defined hitting strategies; The highest priority target hitting strategy is selected from all the target hitting strategies, and the judgment result corresponding to the highest priority target hitting strategy is used as the final judgment result for the driver.

6. The method according to claim 1, characterized in that, The method further includes: Determine whether the final determination indicates that the driver is at fault; If the final determination indicates that the driver is at fault, determine the corresponding appeal label for the driver; If the appeal tag is a full appeal, output the corresponding appeal steps, and detect in real time whether the driver has filed an appeal for the order; If the appeal tag is not a full appeal, output the corresponding appeal pop-up window, and detect in real time whether there is an appeal operation by the driver for the order; If such an order exists, it will be sent to a human review panel.

7. A liability determination system, characterized in that, The system includes: The acquisition module is used to acquire orders and identify the drivers and their information associated with the orders. The preliminary liability determination module is used to obtain a pre-built driver whitelist, and to make a preliminary liability determination on the driver based on the driver whitelist and the driver information to obtain a preliminary determination result; wherein, the driver whitelist is constructed by the driver whitelist construction module based on a pre-built driver liability probability prediction model and historical order data; An algorithm tag determination module is used to, if the preliminary judgment result indicates that the driver is responsible, obtain the order information of the order, and determine the algorithm tag information of each of the pre-built sub-models based on the order information; wherein, the algorithm tag information is empty, or includes at least one algorithm tag; The final responsibility determination module is used to determine multiple hit strategies based on each of the algorithm tags and other algorithm tags, and to make a final responsibility determination on the driver based on each of the hit strategies, so as to obtain the final determination result.

8. The system according to claim 7, characterized in that, The driver whitelist construction module is specifically used for: Obtain historical order data, wherein the historical order data includes multiple historical orders; For each historical order, obtain historical order information related to the historical order, wherein the historical order information includes historical driver information of historical drivers, historical user information of historical users, and the actual probability of the historical driver being at fault. The historical driver information and the historical user information are input into a pre-built driver liability probability prediction model, which then makes a prediction based on the historical driver information and the historical user information to obtain the historical driver's liability probability. The driver liability probability prediction model is trained on a driver liability probability prediction model to be trained based on the historical driver information, historical user information, and the actual liability probability of the historical driver related to each historical order. A driver whitelist is constructed based on the driver IDs of historical drivers whose probability of being at fault is greater than a preset probability.

9. An electronic device, characterized in that, include: A processor and a memory are connected via a bus; wherein the processor is used to call and execute a program stored in the memory; The memory is used to store a program for implementing the liability determination method as described in any one of claims 1-6.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions for performing the liability determination method as described in any one of claims 1-6.

Images