Freight order abnormity processing method and device, electronic equipment and storage medium

Abstract

The invention provides a freight order abnormity processing method and device, electronic equipment and a storage medium. The method comprises the following steps: receiving an order settlement request; extracting N pieces of to-be-settled order data corresponding to the order settlement request and user data; using a pre-constructed risk identification model to identify the N to-be-settled order data and the user data to calculate a risk difference to obtain an order risk result, the order risk result being composed of M feature anomaly scores, M being a positive integer greater than or equal to 2; and when the order risk result is represented as an abnormal state, constructing an exception handling scheme based on the risk type of the order risk result, the exception handling scheme being used for executing a corresponding risk control measure. In the embodiment of the invention, through the to-be-settled order data and the user data, accurate prediction and control of the withdrawal abnormity are automatically carried out on the order settlement request, so that the abnormity identification capability is improved, and the user experience can be improved.

Description

Technical Field

[0001] The present invention relates to the field of risk control data processing technology, and in particular to a method, device, electronic device and storage medium for handling abnormal freight orders. Background Art

[0002] Currently, manual exception systems are often used to quickly determine the risk of applications, and the applications are automatically assigned to reviewers for manual review to determine whether there are any abnormalities in the withdrawal application. Since the review through the above method requires high operating costs and cannot detect complex and hidden identifications in a timely manner, it leads to insufficient abnormality identification capabilities. Summary of the invention

[0003] In view of this, an embodiment of the present invention provides a method, device, electronic device and storage medium for handling freight order exceptions to solve the problem of insufficient exception recognition capability in the prior art.

[0004] To achieve the above objectives, the embodiments of the present invention provide the following technical solutions:

[0005] A first aspect of an embodiment of the present invention shows a method for handling freight order exceptions, the method comprising:

[0006] Receive order settlement request;

[0007] Extracting N to-be-settled order data and user data corresponding to the order settlement request, where N is a positive integer greater than or equal to 1;

[0008] Using a pre-built risk identification model, the N to-be-settled order data and the user data are identified to calculate risk differences, and an order risk result is obtained, wherein the order risk result is composed of M feature anomaly scores, where M is a positive integer greater than or equal to 2;

[0009] When the order risk result is characterized as an abnormal state, an abnormality handling plan is constructed based on the risk type of the order risk result, and the abnormality handling plan is used to execute corresponding risk control measures.

[0010] Optionally, using a pre-built risk identification model to identify the N to-be-settled order data and the user data to calculate risk differences and obtain order risk results, including:

[0011] Extracting N to-be-settled order data and M features of the user data, where M is a positive integer greater than or equal to 2;

[0012] Determine the path length of each feature using a feature tree in a pre-built risk identification model, wherein the risk identification model is built based on historical order data and historical user data;

[0013] The feature anomaly score corresponding to the path length of each feature is calculated using the pre-built risk identification model;

[0014] The pre-built risk identification model is used to take the feature anomaly scores corresponding to the M features as the order risk results.

[0015] Optionally, the process of building a risk identification model based on historical order data and historical user data includes:

[0016] Collect historical order data and historical user data;

[0017] Preprocessing the historical order data and the historical user data to obtain processed freight order data and user behavior data of the historical freight orders;

[0018] The processed historical order data and historical user data are used as a sample set;

[0019] Constructing a preset number of feature trees based on the sample set;

[0020] The preset number of feature trees and the first feature anomaly score calculation method are combined into a risk identification model.

[0021] Optionally, constructing a preset number of feature trees based on the sample set includes:

[0022] Randomly select a preset number of sample subsets from the sample set;

[0023] For each sample subset, the sample subset is used as a root node;

[0024] Dividing the sample subsets based on the features in the root node, and using the divided sample subsets as child nodes of the root node;

[0025] Determining whether the number of features in each of the new nodes is 1, determining whether a constructed tree is obtained;

[0026] A corresponding tree is constructed based on each of the sample subsets to obtain a preset number of feature trees.

[0027] Optionally, the order risk result is characterized as an abnormal state, including:

[0028] For each feature, determining whether the feature is a risk feature based on the feature anomaly score corresponding to the feature in the order risk result;

[0029] If yes, counting the first quantity of the risk feature;

[0030] Determining whether the first number is greater than a preset number;

[0031] If it is greater, the order risk result is determined to be abnormal.

[0032] Optionally, build an exception handling plan based on the risk type of the order risk result, including:

[0033] determining a risk type of the order risk result based on the first amount;

[0034] Find the risk control measures corresponding to the risk type of the order risk result;

[0035] Construct an exception handling plan based on the risk control measures.

[0036] Optionally, a pre-built risk identification model is used to process the N to-be-settled order data and the user data to generate an order risk result, including:

[0037] Extracting N to-be-settled order data and M features of the user data, where M is a positive integer greater than or equal to 1;

[0038] Using a pre-built risk identification model to identify M features and determine the identities of the M features, the risk identification model is constructed by classification based on historical order data and historical user data;

[0039] The identifications of the M features are combined to generate an order risk result.

[0040] A second aspect of an embodiment of the present invention shows a device for handling abnormalities in a freight order, the device comprising:

[0041] A receiving unit, used for receiving an order settlement request;

[0042] An extraction unit, used to extract N to-be-settled order data and user data corresponding to the order settlement request, where N is a positive integer greater than or equal to 1;

[0043] A pre-built risk identification model is used to identify the N to-be-settled order data and the user data using the pre-built risk identification model to calculate risk differences and obtain order risk results;

[0044] A processing unit is used to construct an exception handling plan based on the risk type of the order risk result when the order risk result is characterized as an abnormal state, and the exception handling plan is used to execute corresponding risk control measures.

[0045] A third aspect of an embodiment of the present invention shows an electronic device, which is used to run a program, wherein when the program is run, the freight order exception processing device shown in the first aspect of the embodiment of the present invention is executed.

[0046] The fourth aspect of an embodiment of the present invention shows a storage medium, which includes a storage program, wherein when the program is running, the device where the storage medium is located is controlled to execute the freight order exception processing device as shown in the first aspect of the embodiment of the present invention.

[0047] Based on the above-mentioned embodiment of the present invention, a method, device, electronic device and storage medium for handling freight order exceptions are provided, the method comprising: receiving an order settlement request; extracting N pending order data and user data corresponding to the order settlement request, wherein N is a positive integer greater than or equal to 1; using a pre-built risk identification model to identify the N pending order data and the user data to calculate the risk difference, and obtain an order risk result, wherein the order risk result is composed of M feature anomaly scores, wherein M is a positive integer greater than or equal to 2; when the order risk result is characterized as an abnormal state, an exception handling scheme is constructed based on the risk type of the order risk result, and the exception handling scheme is used to execute corresponding risk control measures. In an embodiment of the present invention, the withdrawal anomalies of the order settlement request are automatically and accurately predicted and controlled through the pending order data and user data, thereby improving the anomaly identification capability and improving the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] 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 embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0049] Figure 1 A schematic diagram of interaction between a mobile terminal and a server according to an embodiment of the present invention;

[0050] Figure 2 A flow chart of a method for handling freight order exceptions according to an embodiment of the present invention;

[0051] Figure 3 A schematic diagram of an exception handling process shown in an embodiment of the present invention;

[0052] Figure 4 This is a schematic diagram of the architecture of cash withdrawal for freight orders according to an embodiment of the present invention;

[0053] Figure 5 The present invention is a schematic diagram of the structure of a device for handling abnormal freight orders according to an embodiment of the present invention. DETAILED DESCRIPTION

[0054] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0055] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0056] It should be noted that the descriptions of "first", "second", etc. in the present invention are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in the field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0057] In this application, the terms "comprises", "comprising" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device comprising 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, an element defined by the sentence "comprising a ..." does not exclude the presence of other identical elements in the process, method, article or device comprising the element.

[0058] As can be seen from the background technology, currently, the manual risk control system optimized by the rule engine can automatically execute risk assessment rules, quickly and accurately determine the risk level of the application, and automatically assign auditors for manual review, thereby improving the efficiency and accuracy of risk control processing. At the same time, the rule engine can be flexibly adjusted and optimized according to actual conditions to adapt to different business scenarios and risk situations.

[0059] However, the current manual risk control system optimized by rule engines still has some shortcomings. For example, the rule writing of rule engines requires certain technical capabilities and experience, which may be difficult for non-technical personnel; at the same time, the rule base of the rule engine needs to be constantly updated and maintained to ensure its accuracy and adaptability, which leads to insufficient anomaly recognition capabilities and cannot meet the ever-changing business needs and risk situations.

[0060] An embodiment of the present invention provides a new method for handling freight order exceptions. In a scenario where an online car-hailing driver accepts orders and deposits or withdraws cash, different recognition models are shown to identify the first order corresponding to the order settlement request, and anomalies are identified based on the freight order data and user behavior data of the first order, so as to efficiently detect abnormal behaviors, thereby improving user experience and meeting ever-changing business needs and risk situations.

[0061] See also Figure 1 , is a schematic diagram of interaction between a mobile terminal and a server according to an embodiment of the present invention, wherein the method comprises:

[0062] The mobile terminal 10 is wirelessly connected to the server 20 .

[0063] The number of the mobile terminals is multiple.

[0064] The process of implementing the freight order exception handling based on the above-mentioned architecture includes:

[0065] When the server 20 receives an order settlement request initiated by a user based on the mobile terminal 10, it extracts N to-be-settled order data and user data corresponding to the order settlement request based on the order settlement request, where N is a positive integer greater than or equal to 1; uses a pre-built risk identification model to identify the N to-be-settled order data and the user data to calculate the risk difference, and obtains an order risk result, where the order risk result is composed of M feature anomaly scores, where M is a positive integer greater than or equal to 2; when the order risk result is characterized as an abnormal state, an exception handling plan is constructed based on the risk type of the order risk result, and the exception handling plan is used to execute corresponding risk control measures.

[0066] In an embodiment of the present invention, withdrawal anomalies are automatically and accurately predicted and controlled based on the freight order data and the user behavior data, thereby improving the ability to identify anomalies and enhancing the user experience.

[0067] See also Figure 2 , is a flow chart of a method for handling freight order exceptions according to an embodiment of the present invention, the method comprising:

[0068] Step S201: receiving an order settlement request.

[0069] Step S202: extracting N to-be-settled order data and user data corresponding to the order settlement request, where N is a positive integer greater than or equal to 1;

[0070] Optionally, the freight application in the cargo owner's mobile terminal submits the order to be shipped and pays the corresponding transportation amount. When the server confirms that it has received the order to be shipped, it publishes it so that the driver can accept the order.

[0071] After the driver finishes transporting an order, the server will write the freight amount corresponding to the order into the driver's corresponding account, so that the driver can subsequently enter the freight amount he wants to withdraw through the freight application in the mobile terminal to withdraw the order. The mobile terminal will generate a corresponding order settlement request and send it to the server.

[0072] That is to say, the order settlement request carries the freight amount.

[0073] It should be noted that the user shown in the embodiment of the present invention may be a driver.

[0074] In the process of implementing step S201 and step S202, first, when the server receives an order settlement request initiated by a user on a mobile terminal, it determines whether the freight amount carried in the order settlement request is greater than the withdrawable amount in the driver's account. If so, a prompt message is output to prompt the driver to re-enter the freight amount he wants to withdraw. If not, it indicates that the freight amount can be withdrawn.

[0075] Next, all orders in the driver's account for which the freight amount has not been withdrawn and for which the freight transportation has been completed, i.e., first orders, are traversed to determine the freight amount of each first order.

[0076] If the freight amount to be withdrawn carried in the order settlement request is equal to the cumulative value of the freight amounts of all first orders, extract N to-be-settled order data and user data corresponding to the order settlement request, where N is equal to the number of all first orders;

[0077] If the freight amount to be withdrawn carried in the order settlement request is equal to the cumulative value of the freight amounts of part of the first order, that is, N orders to be settled, then obtain N orders to be settled data and user data, where N is a positive integer greater than or equal to 1;

[0078] For example: Driver A's Account 1 contains orders for which the freight amount has not been withdrawn and the freight transportation has been completed, namely, the first order A and the first order B.

[0079] Among them, the freight amount of the first order A and the first order B is both 1,000 yuan. If the freight amount to be withdrawn 1,000 carried in the order settlement request is equal to the freight amount of the first order A, and is also equal to the freight amount of the first order 2, then select any first order, such as the first order A, and then obtain part of the order data to be settled of the first order A, as well as the user data.

[0080] It should be noted that user data includes withdrawal frequency, withdrawal time, withdrawal location, and historical withdrawal records.

[0081] For example, on an online ride-hailing platform, the server collects user data such as the driver's withdrawal record, withdrawal frequency, withdrawal time, withdrawal location, historical withdrawal record, score, historical violation record, etc.

[0082] User data may also include the driver's social network characteristics, such as the frequency of interaction with other drivers.

[0083] The order data to be settled refers to the data of one or more freight orders, specifically including the freight-related data such as the freight transported goods, the freight starting location, the freight destination, the vehicle route corresponding to the freight order, the mileage and transportation time, etc.

[0084] Step S203: using a pre-built risk identification model to identify the N to-be-settled order data and the user data to calculate risk differences and obtain order risk results.

[0085] The order risk result is composed of M feature anomaly scores, where M is a positive integer greater than or equal to 2.

[0086] It should be noted that there are multiple ways to implement the process of step S202.

[0087] The first implementation method, such as Figure 3 As shown, the following steps are included:

[0088] Step S301: extracting N to-be-settled order data and M features of the user data, where M is a positive integer greater than or equal to 2;

[0089] Step S302: Determine the path length of each feature using a feature tree in a pre-constructed risk identification model, wherein the risk identification model is constructed based on historical order data and historical user data;

[0090] Specifically, the process of building a risk identification model based on historical order data and historical user data includes:

[0091] Step S11: Collect historical order data and historical user data;

[0092] In the specific process of implementing step S11, order data and historical user data of historical freight orders in the past historical time are collected.

[0093] It should be noted that there are multiple historical freight orders.

[0094] The historical time is set by technicians based on experience or experiments, and the embodiment of the present invention does not limit this.

[0095] Step S12: pre-processing the historical order data and historical user data to obtain processed freight order data and user behavior data of the historical freight order;

[0096] In the specific implementation of step S12, data standardization or normalization is performed on the historical order data and historical user data of the historical freight orders to obtain processed historical order data and historical user data.

[0097] Step S13: taking the processed historical order data and historical user data as a sample set;

[0098] Among them, the historical order data and historical user data of each historical freight order are one piece of data in the sample set.

[0099] Step S14: constructing a preset number of feature trees based on the sample set;

[0100] The specific process of implementing step S14 includes the following steps:

[0101] Step S21: randomly extracting a preset number of sample subsets from the sample set;

[0102] In the specific implementation of step S21, first, multiple pieces of data are repeatedly randomly extracted from the sample set, and the multiple pieces of data extracted each time are used as a sample subset to obtain a preset number of sample subsets.

[0103] The number of repetitions is equal to the preset number, that is, the number of sample subsets is the preset number.

[0104] Step S22: for each sample subset, taking the sample subset as a root node;

[0105] Step S23: dividing the sample subsets based on the features in the root node, and using the divided sample subsets as child nodes of the root node;

[0106] Step S24: Determine whether the number of features in each of the new nodes is 1. If it is determined that the number of features in each of the new nodes is 1, execute step S25.

[0107] Step S25: Determine that the constructed tree is obtained;

[0108] In the process of specifically implementing steps S22 to S25, first, the sample subset is taken as the root node, and for the root node, a feature x1 and a random split value t of the feature x1 are randomly selected from the sample subset; each other feature xj in the sample subset is divided according to the random split value t, and it is determined whether the split value tj corresponding to the feature xj is greater than the random split value t, if greater, the feature xj is written into the first subset, and if less than or equal to, the feature xj is written into the second subset; then, the first subset and the second subset are respectively taken as child nodes of the root node, that is, new nodes; then, for the new node, a feature x2 and a random split value t of the feature x2 are randomly selected from the new node to divide the subset in the new node according to the random split value t of the feature x2, so as to repeat the above processing process, thereby dividing the new node.

[0109] It should be noted that the process of constructing each new node is the same as the process of dividing the root node mentioned above, and they can refer to each other.

[0110] Finally, corresponding to each new node, check whether the number of features in the subset of each new node is 1. If so, determine that the stopping condition is met and obtain the constructed tree. If not, continue to divide the new node to obtain the child nodes of the new node.

[0111] It should be noted that each tree is an isolated tree.

[0112] In the embodiment of the present invention, historical order data and historical user data are partitioned by characteristics such as geographic location or time, and processed in parallel on different nodes. Data flows along a preset path, from preprocessing to calculation to output, and each node decomposes the task into subtasks to achieve parallel processing, thereby significantly improving the computing efficiency of online traffic and the ability to detect anomalies.

[0113] Optionally, it also includes: after the node division is completed, first obtain the depth of the tree in the current tree construction process, determine whether the depth of the tree reaches a preset maximum value, and when it is determined that the depth of the tree reaches the preset maximum value, stop further division to obtain a completed tree.

[0114] Based on the above process, a tree corresponding to each sample subset is constructed to obtain a preset number of trees.

[0115] Step S26: construct a corresponding tree based on each of the sample subsets to obtain a preset number of feature trees.

[0116] In the specific implementation of step S26, features are extracted from historical order data and historical user data. For each feature x, each node in the tree under the risk identification model is traversed to determine which leaf node the feature x is in, and the position of the leaf node is recorded.

[0117] It should be noted that a leaf node refers to the terminal node of the tree.

[0118] The number of features in the historical order data and the historical user data is n, where n is a positive integer greater than or equal to 2.

[0119] Step S15: The preset number of feature trees and the first feature anomaly score calculation method are combined into a risk identification model.

[0120] In the process of implementing step S301 and step S302, N to-be-settled order data and M features of the user data are extracted, and the positions of the M features in each feature tree of the risk identification model are determined; the path length from the position of its leaf node to the root node of the corresponding tree is calculated.

[0121] Step S303: Calculate the feature anomaly score corresponding to the path length of each feature using the pre-built risk identification model;

[0122] In the specific implementation process of step S303, the average path length E(h(x)) of the feature x on all trees is calculated; the average path length E(h(x)) corresponding to each feature is substituted into formula (1) for calculation to obtain the withdrawal anomaly score; and the withdrawal anomaly score corresponding to each feature is used as the processing result.

[0123] Formula (1):

[0124]

[0125] Wherein, c(n) is a normalization factor used to standardize the path length, which depends on the number of features n in the freight order data and the user behavior data, and can be calculated by the following formula (2).

[0126] Formula (2):

[0127]

[0128] Where H(i) is the harmonic number, which can be approximately calculated using the following formula (3).

[0129] Formula (3):

[0130] H(i)=ln(i)+γ(3)

[0131] Wherein, γ is Euler's constant, which is approximately equal to 0.5772156649. From formula (2), we can know that i can be n-1.

[0132] Step S304: using the pre-built risk identification model to take the feature anomaly scores corresponding to the M features as the order risk results.

[0133] The second implementation method includes the following steps:

[0134] Step S31: extracting N to-be-settled order data and M features of the user data, where M is a positive integer greater than or equal to 1;

[0135] Step S32: using a pre-built risk identification model to identify M features and determine the identifiers of the M features, wherein the risk identification model is constructed by classifying historical order data and historical user data;

[0136] It should be noted that the process of classifying and constructing a risk identification model based on historical order data and historical user data includes:

[0137] Step S41: Collect historical order data and historical user data of historical freight orders.

[0138] In the specific process of implementing step S41, historical order data and historical user data of historical freight orders in the past historical time are collected.

[0139] It should be noted that there are multiple historical freight orders.

[0140] The historical time is set by technicians based on experience or experiments, and the embodiment of the present invention does not limit this.

[0141] Step S42: pre-processing the historical order data and historical user data to obtain processed freight order data and user behavior data of the historical freight order;

[0142] In the specific implementation of step S42, data standardization or normalization is performed on the historical order data and historical user data of the historical freight orders to obtain processed historical order data and historical user data.

[0143] Step S43: marking abnormal data and normal data in the historical order data and historical user data of the historical freight order;

[0144] Step S44: taking the marked historical order data and historical user data as a sample set;

[0145] Step S45: Use the sample set to train the initial model to obtain a risk identification model.

[0146] In the specific implementation of steps S43 to S45, the first feature related to the order-brushing behavior in the historical order data and historical user data of the historical freight orders is marked, such as the number of orders completed in a short period of time, the consistency of the order and the GPS track, the abnormal pattern of customer evaluation, the abnormal pattern of cash withdrawal behavior, etc.; and the second feature related to the normal cash withdrawal behavior in the historical order data and historical user data is marked; the first feature and the second feature are converted into corresponding feature vectors, and used as a sample set;

[0147] Next, the sample set is divided into two parts, one as a training set and the other as a test set. The training set is used to perform classification training on the initial model to obtain a trained initial model. The trained initial model is tested using the test set. If the output recognition result is consistent with the label of the test set, the current initial model is used as a risk identification model.

[0148] It should be noted that the initial model is a support vector machine or a deep learning model.

[0149] In the specific implementation of step S32, the pre-built risk identification model is used to identify M features, and identifiers of the M features are determined, where the identifiers are used to indicate whether the features are abnormal data.

[0150] Step S33: Combining the identifiers of the M features to generate an order risk result.

[0151] In the specific implementation of step S33, the identification results of the M features identified as abnormal data or normal are combined to generate an order risk result.

[0152] It should be noted that the order risk results include indicators indicating abnormal data and indicators indicating normal data.

[0153] Data anomalies refer to order data in which drivers engage in behaviors such as false orders.

[0154] Step S204: Determine whether the order risk result is characterized as an abnormal state. If so, execute step S205.

[0155] It should be noted that there are multiple ways to implement the process of step S204.

[0156] If the order risk result includes the feature anomaly score corresponding to each feature, the specific implementation method includes the following steps:

[0157] Step S51: for each feature, determine whether the feature is a risk feature based on the feature anomaly score corresponding to the feature in the order risk result. If so, execute step S52. If not,

[0158] In the process of implementing step S51, the withdrawal anomaly score of each feature in the order risk result is determined to be greater than the preset threshold. If there is a withdrawal anomaly score greater than the preset threshold, the corresponding feature is used as a risk feature and recorded, and then step S52 is executed. If not, the corresponding feature is used as a normal feature and the feature is discarded.

[0159] It should be noted that the preset threshold is set by the technicians in advance based on multiple experiments or experience. For example, if the withdrawal anomaly score s(x,n) is greater than the preset threshold 0.5, it means that feature x is abnormal.

[0160] Step S52: Counting a first quantity of the risk feature;

[0161] In the specific implementation of step S52, a first number of risk features among the M features is counted;

[0162] Step S53: Determine whether the first quantity is greater than or equal to a preset quantity. If so, execute step S54. If not, it means that the order settlement request is normal, and execute the withdrawal operation corresponding to the order settlement request.

[0163] It should be noted that the preset number is set by technicians after multiple experiments or experience.

[0164] If the first quantity is greater than or equal to the preset quantity, it means that the first order may have abnormalities such as an abnormally fast average speed, an abnormally high login frequency, an abnormally low evaluation score, or an abnormal withdrawal amount.

[0165] Optionally, after executing step S3, further confirmation can be performed, that is, checking again whether there are any abnormalities in the order to be settled. The verification can be carried out by other means. At this time, new features are added, such as driving video or call records on the driver's side, and they are input into another risk identification model together with previous data for identification.

[0166] Optionally, the results of the re-review or manual review are fed back into the model for retraining and optimization of the model.

[0167] Furthermore, the driver behavior database is updated to record the confirmed order-brushing behaviors and corresponding characteristics.

[0168] Step S54: Determine whether the order risk result is characterized as an abnormal state.

[0169] If the order risk result includes the identification of M features, the specific implementation method includes the following steps:

[0170] Step S61: counting a second number of identifiers indicating abnormal data in the order risk result;

[0171] In the specific implementation of step S61, the number of features identified as abnormal data among the M features is counted and used as the second number;

[0172] Step S62: Determine whether the second quantity is greater than a preset quantity. If so, execute step S63. If not, it indicates that the order settlement request is normal, and execute the withdrawal operation corresponding to the order settlement request.

[0173] Step S63: Determine that the order risk result is characterized as an abnormal state.

[0174] The present invention constructs a corresponding risk identification model through machine learning and deep learning technology, and then calculates the identification of each feature in the order settlement request based on the risk identification model, and analyzes the user's withdrawal behavior based on whether the identification indicates an abnormality, thereby evaluating the user's withdrawal risk.

[0175] Step S205: Construct an exception handling plan based on the risk type of the order risk result.

[0176] The specific implementation process of step S205 includes the following steps:

[0177] Step S71: determining the risk type of the order risk result based on the first quantity;

[0178] Step S72: searching for risk control measures corresponding to the risk type of the order risk result;

[0179] Step S73: Construct an exception handling plan based on the risk control measures.

[0180] In the specific implementation of step S71 to step S73, first, the ratio of the first number to the total number of features M is calculated, and the corresponding risk type is searched according to the ratio.

[0181] Among them, the corresponding relationship between different ratios and risk types is preset, and the higher the ratio, the more important the corresponding risk type is.

[0182] Next, searching for a correspondence between pre-stored risk control measures and risk types, and determining a risk control measure corresponding to the risk type of the order risk result;

[0183] Among them, the correspondence between pre-stored risk control measures and risk types.

[0184] Then, an exception handling plan is constructed based on the risk control measures. For example, if the risk control measure is freeze control, the exception handling plan may be to automatically trigger an early warning and freeze the account corresponding to the order settlement request. The freeze control belongs to the control measure.

[0185] Optionally, for situations where the risk is higher but cannot be confirmed, increase the monitoring intensity of the account and limit the frequency and amount of withdrawals.

[0186] The present invention forms a corresponding risk identification model through multiple trees, then identifies the order settlement request based on the risk identification model, and determines whether the order settlement request is abnormal based on the withdrawal anomaly score of each feature of the order risk result, thereby improving the anomaly identification capability, that is, effectively identifying and preventing the driver's false order-brushing behavior, protecting the interests of the platform and consumers, and improving the overall quality of the service.

[0187] Optionally, the specific implementation process based on the above steps S201 to S204 can also be described by a schematic diagram, such as Figure 4 shown.

[0188] Figure 4 In the data collection module, the payment behavior data corresponding to the historical freight orders are first collected, and the payment behavior data is cleaned and preprocessed through the data processing module to build a risk identification model. Then, the risk assessment model, i.e., the risk identification model, is used to conduct risk assessment on the current order settlement request to obtain the order risk result; if the order risk result is characterized as an abnormal state, an early warning is issued through the risk early warning model; and according to the processing result, i.e., the assessment result, risk control measures are taken through the risk control module.

[0189] Then the data feedback module feeds back the risk control processing results received from the risk control module to the data collection module.

[0190] In an embodiment of the present invention, in a scenario where an online car-hailing driver accepts orders and deposits or withdraws cash, a pre-constructed risk identification model is used to identify order data and user data in an order settlement request to obtain an order risk result. When the order risk result is characterized as an abnormal state, an exception handling plan is constructed based on the risk type of the order risk result. The exception handling plan is used to execute corresponding risk control measures, thereby improving the user experience.

[0191] Based on the freight order exception processing method shown in the above embodiment of the present invention, correspondingly, the embodiment of the present invention shows a freight order exception processing device, such as Figure 5 As shown, the device comprises:

[0192] Receiving unit 501, used to receive an order settlement request;

[0193] An extraction unit 502 is used to extract N to-be-settled order data and user data corresponding to the order settlement request, where N is a positive integer greater than or equal to 1;

[0194] A pre-built risk identification model 503 is used to identify the N to-be-settled order data and the user data using the pre-built risk identification model to calculate risk differences and obtain order risk results;

[0195] The processing unit 504 is used to construct an abnormal processing scheme based on the risk type of the order risk result when the order risk result is characterized as an abnormal state, and the abnormal processing scheme is used to execute corresponding risk control measures.

[0196] The specific principles and execution processes of each unit in the freight order exception processing device disclosed in the above embodiment of the present invention are the same as the corresponding contents in the freight order exception processing method provided in the above embodiment of the present invention. Please refer to the corresponding parts of the freight order exception processing method disclosed in the above embodiment of the present invention, and no further details will be given here.

[0197] In an embodiment of the present invention, in a scenario where an online car-hailing driver accepts orders and deposits or withdraws cash, a pre-constructed risk identification model is used to identify order data and user data in an order settlement request to obtain an order risk result. When the order risk result is characterized as an abnormal state, an exception handling plan is constructed based on the risk type of the order risk result. The exception handling plan is used to execute corresponding risk control measures, thereby improving the user experience.

[0198] Optionally, based on the apparatus for handling freight order anomalies shown in the above embodiment of the present invention, the pre-constructed risk identification model 503 is specifically used for:

[0199] Extracting N to-be-settled order data and M features of the user data, where M is a positive integer greater than or equal to 2;

[0200] Determine the path length of each feature using a feature tree in a pre-built risk identification model, wherein the risk identification model is built based on historical order data and historical user data;

[0201] The feature anomaly score corresponding to the path length of each feature is calculated using the pre-built risk identification model;

[0202] The pre-built risk identification model is used to take the feature anomaly scores corresponding to the M features as the order risk results.

[0203] The process of building a risk identification model based on historical order data and historical user data includes:

[0204] Collect historical order data and historical user data;

[0205] Preprocessing the historical order data and the historical user data to obtain processed freight order data and user behavior data of the historical freight orders;

[0206] The processed historical order data and historical user data are used as a sample set;

[0207] Constructing a preset number of feature trees based on the sample set;

[0208] The preset number of feature trees and the first feature anomaly score calculation method are combined into a risk identification model.

[0209] Wherein, constructing a preset number of feature trees based on the sample set includes:

[0210] Randomly select a preset number of sample subsets from the sample set;

[0211] For each sample subset, the sample subset is used as a root node;

[0212] Dividing the sample subsets based on the features in the root node, and using the divided sample subsets as child nodes of the root node;

[0213] Determining whether the number of features in each of the new nodes is 1, determining whether a constructed tree is obtained;

[0214] A corresponding tree is constructed based on each of the sample subsets to obtain a preset number of feature trees.

[0215] Optionally, the processing unit 504 for characterizing the order risk result as an abnormal state is specifically configured to:

[0216] For each feature, determining whether the feature is a risk feature based on the feature anomaly score corresponding to the feature in the order risk result;

[0217] If yes, counting the first quantity of the risk feature;

[0218] Determining whether the first number is greater than a preset number;

[0219] If it is greater, the order risk result is determined to be abnormal.

[0220] Optionally, based on the freight order exception processing device shown in the above embodiment of the present invention, a processing unit 504 for constructing an exception processing solution based on the risk type of the order risk result is specifically used to:

[0221] determining a risk type of the order risk result based on the first amount;

[0222] Find the risk control measures corresponding to the risk type of the order risk result;

[0223] Construct an exception handling plan based on the risk control measures.

[0224] Optionally, based on the apparatus for handling freight order anomalies shown in the above embodiment of the present invention, the pre-constructed risk identification model 503 is also used for:

[0225] Extracting N to-be-settled order data and M features of the user data, where M is a positive integer greater than or equal to 1;

[0226] Using a pre-built risk identification model to identify M features and determine the identities of the M features, the risk identification model is constructed by classification based on historical order data and historical user data;

[0227] The identifications of the M features are combined to generate an order risk result.

[0228] An embodiment of the present application provides an electronic device, which includes a processor and a memory, the memory being used for program code and data for freight order exception processing, and the processor being used to call program instructions in the memory to execute the steps shown in the freight order exception processing method in the above embodiment.

[0229] An embodiment of the present application provides a storage medium, which includes a storage program, wherein when the program is running, the device where the storage medium is located is controlled to execute the freight order exception processing method shown in the above embodiment.

[0230] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can refer to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system or system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can refer to the partial description of the method embodiment. The system and system embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without creative work.

[0231] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0232] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one 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 present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for handling abnormal freight orders, characterized in that: The method comprises: Receive order settlement request; Extracting N to-be-settled order data and user data corresponding to the order settlement request, where N is a positive integer greater than or equal to 1; Using a pre-built risk identification model, the N to-be-settled order data and the user data are identified to calculate risk differences, and an order risk result is obtained, wherein the order risk result is composed of M feature anomaly scores, where M is a positive integer greater than or equal to 2; When the order risk result is characterized as an abnormal state, an abnormality handling plan is constructed based on the risk type of the order risk result, and the abnormality handling plan is used to execute corresponding risk control measures.

2. The method according to claim 1, characterized in that: The risk identification model constructed in advance is used to identify the N pending order data and the user data to calculate the risk difference, and obtain the order risk result, including: Extracting N to-be-settled order data and M features of the user data, where M is a positive integer greater than or equal to 2; Determine the path length of each feature using a feature tree in a pre-built risk identification model, wherein the risk identification model is built based on historical order data and historical user data; The feature anomaly score corresponding to the path length of each feature is calculated using the pre-built risk identification model; The pre-built risk identification model is used to take the feature anomaly scores corresponding to the M features as the order risk results.

3. The method according to claim 1, characterized in that The process of building a risk identification model based on historical order data and historical user data includes: Collect historical order data and historical user data; Preprocessing the historical order data and the historical user data to obtain processed freight order data and user behavior data of the historical freight orders; The processed historical order data and historical user data are used as a sample set; Constructing a preset number of feature trees based on the sample set; The preset number of feature trees and the first feature anomaly score calculation method are combined into a risk identification model.

4. The method according to claim 3, characterized in that Constructing a preset number of feature trees based on the sample set, including: Randomly select a preset number of sample subsets from the sample set; For each sample subset, the sample subset is used as a root node; Dividing the sample subsets based on the features in the root node, and using the divided sample subsets as child nodes of the root node; By determining whether the number of features in each new node is 1, it is determined that a completed tree is obtained; A corresponding tree is constructed based on each of the sample subsets to obtain a preset number of feature trees.

5. The method according to claim 1, characterized in that The order risk results are characterized as abnormal states, including: For each feature, determining whether the feature is a risk feature based on the feature anomaly score corresponding to the feature in the order risk result; If yes, counting the first quantity of the risk feature; Determining whether the first number is greater than a preset number; If it is greater, the order risk result is determined to be abnormal.

6. The method according to claim 5, characterized in that Build exception handling solutions based on the risk type of order risk results, including: determining a risk type of the order risk result based on the first amount; Find the risk control measures corresponding to the risk type of the order risk result; Construct an exception handling plan based on the risk control measures.

7. The method according to claim 1, characterized in that The risk identification model constructed in advance is used to identify the N pending order data and the user data to calculate the risk difference, and obtain the order risk result, including: Extracting N to-be-settled order data and M features of the user data, where M is a positive integer greater than or equal to 1; Using a pre-built risk identification model to identify M features and determine the identities of the M features, the risk identification model is constructed by classification based on historical order data and historical user data; The identifications of the M features are combined to generate an order risk result.

8. A freight order exception processing device, characterized in that: The device comprises: A receiving unit, used for receiving an order settlement request; An extraction unit, used to extract N to-be-settled order data and user data corresponding to the order settlement request, where N is a positive integer greater than or equal to 1; A pre-built risk identification model is used to identify the N to-be-settled order data and the user data using the pre-built risk identification model to calculate risk differences and obtain order risk results; A processing unit is used to construct an exception handling plan based on the risk type of the order risk result when the order risk result is characterized as an abnormal state, and the exception handling plan is used to execute corresponding risk control measures.

9. An electronic device, characterized in that: The electronic device is used to run a program, wherein the program, when running, executes the freight order exception processing device as described in any one of claims 1-7.

10. A storage medium, characterized in that: The storage medium includes a storage program, wherein when the program is running, the device where the storage medium is located is controlled to execute the freight order exception processing device as described in any one of claims 1-7.

Images