A method for abnormal early warning control of network freight advance payment

Abstract

The present invention discloses a method for abnormal early warning control of network freight freight advance payment, which belongs to the field of logistics information technology, including: first, completing the classification definition of abnormal types in the project abnormal monitoring configuration stage, constructing a judgment model for judging business false class abnormalities, and setting corresponding algorithm rules related to abnormal monitoring, calculating the risk coefficient according to the abnormal records of the corresponding logistics company in the advance payment process, and incentivizing the logistics company according to the incentive measures determined thereby; then performing project abnormal monitoring execution, timing abnormal definition and abnormal algorithm rule execution monitoring, and actively discovering abnormalities; once an abnormality is discovered, a message is immediately pushed to the corresponding personnel after recording; then corresponding processing is performed according to the abnormal type. The present invention promptly discovers abnormalities that will cause major or serious risks to the platform advance payment of freight, and can motivate logistics companies to perform logistics processes and operate the system of this platform according to specifications, reducing the difficulty and time of management personnel to investigate.

Description

Technical Field

[0001] The present invention belongs to the field of logistics information technology and relates to a method for abnormal early warning control of network freight advance payment. Background Art

[0002] In recent years, with the rapid development of online freight business, online freight platforms have gradually emerged. Many logistics companies have launched online freight platforms to legally obtain income and expand their own sources of transportation. Due to the expansion of business scale and other special circumstances, logistics companies often lack funds, which requires online freight platforms to have the ability to advance freight and solve the capital needs of logistics companies on the platform.

[0003] Although some online freight platforms currently have the function of freight advance payment, they often do not have the ability to control the abnormalities of freight advance payment, and are unable to promptly discover the risks of the freight advance payment process. This will lead to a greater financial security risk in the freight advance payment process. Once a logistics company withdraws funds from the platform or is unable to repay, it will cause significant losses to the online freight platform. The present invention provides a standardized method for controlling the abnormalities of freight advance payment on an online freight platform. During the business operation process, the business is monitored in all directions, various types of abnormalities are identified and followed up, thereby solving the problem that the existing technology cannot promptly identify the true business situation and repayment ability of the logistics company, resulting in the inability to effectively evaluate and control the risks of freight advance payment. It also solves the problem that the risk monitoring process is easily misled by the logistics company due to non-standard execution and system operation errors in the logistics process, resulting in frequent misidentification of abnormal categories, and it is difficult to eliminate such interference. Summary of the invention

[0004] The purpose of the present invention is to provide a method for abnormal early warning control of freight advance payment on the Internet, which is used to solve the problem that the prior art lacks data information to identify and judge the actual business situation and repayment ability of logistics companies on the Internet freight platform, and the identification results are easily affected by the irregular execution and operation of the logistics companies.

[0005] The method for abnormal early warning control of network freight freight advance payment includes: first, completing the classification definition of abnormal types in the project abnormality monitoring configuration stage, building a judgment model for judging business false type abnormalities, and setting corresponding algorithm rules related to abnormality monitoring, calculating the risk coefficient according to the abnormal records of the corresponding logistics company in the advance payment process, and incentivizing the logistics company according to the incentive measures determined in this way; then executing project abnormality monitoring, timing abnormality definition and abnormal algorithm rule execution monitoring, and actively discovering abnormalities; once an abnormality is found, a message is pushed to the corresponding personnel immediately after recording; then corresponding processing is carried out according to the abnormality type;

[0006] The types of anomalies include at least business false anomalies and capital anomalies. Business false anomalies refer to the discrepancy between the actual logistics data and the waybill contents. Capital anomalies refer to the arrears of funds in the data of capital transactions such as freight and repayment. The algorithm of risk coefficient K is as follows:

[0007]

[0008] In the above formula, the total number of waybills is n, the weight of the business false class anomaly is w1, and as long as any business false class anomaly occurs in a waybill, the business false class anomaly value t i is recorded as 1, otherwise it is recorded as 0; the weight of the fund abnormality class is w2. If any fund abnormality occurs in a waybill, the fund abnormality value s i It is recorded as 1, otherwise it is recorded as 0; w1 and w2 are both between 0-100%, w1+w2=100%.

[0009] Preferably, the judgment of business false anomalies includes: the factors involved in judging whether the business is false are derived from the data collected in the actual logistics process and the data provided by the waybill, the collected data is processed and the indicators are extracted and brought into the judgment model, the probability value is calculated, and the calculation result is compared with the critical value to judge whether the business corresponding to the waybill is false; the judgment model is established according to the logistic regression model as follows:

[0010]

[0011] Among them, Xi (i = 0, 1, 2, ..., 6) represents the indicator variable, X1-X6 represent respectively: the vehicle trajectory does not match the waybill route, the driver does not operate the APP, the waybill has no consignee receipt record, the transportation price is higher than the mean by a certain threshold, the vehicle does not appear at the pick-up and unloading place at the specified time, and the driver does not collect the freight; P is the probability, β i (i=0, 1, 2, ..., 6) represents the regression coefficient, which is the coefficient to be determined; the loss function is defined as L(a, y)=-ylog(a)-(1-y)log(1-a), y represents the true label, and a represents the predicted result; the indicator is entered into the logistic regression model according to backward stepwise regression, and each regression coefficient is determined to finally obtain a determined judgment model.

[0012] Preferably, the actual logistics data includes vehicle trajectories, which are used to identify situations where the vehicle does not appear at the pickup location at the specified time and where the vehicle trajectory does not match the waybill route; the method uses a distance-based outlier recognition model to quickly identify abnormal coordinates, thereby correcting the vehicle trajectory; the outlier recognition method includes:

[0013] (1) First, the driving trajectory of the transport vehicle is segmented;

[0014] (2) Identify outliers in each segment;

[0015] (3) Delete the outliers in each segment and redraw the trajectory.

[0016] Preferably, the step (2) comprises the following specific steps:

[0017] a) Calculate the distance between each point (lonB, latB) in the segment and the starting point (lonA, latA)

[0018]

[0019] Where R is the radius of the earth;

[0020] b) The actual distance traveled is dist2 = s*t, where s is the speed of the vehicle at that point and t is the time difference from that point to the starting point;

[0021] c) If If the value of is greater than the set comparison threshold, the point is considered to be an abnormal point.

[0022] Preferably, the determination of fund anomalies includes: obtaining fund transaction data of each logistics company every day and checking whether there are the following situations:

[0023] A. Delayed payment of freight: The logistics company delays paying the driver freight for more than the set number of days, and the amount exceeds the set amount threshold;

[0024] B. Failure to repay on time: The logistics company fails to repay the loan within the specified time.

[0025] Preferably, the abnormality type also includes operating abnormality, which means that the logistics company's operating conditions continue to be poor over a period of time; the determination of operating abnormality includes: obtaining the capital transaction data of each logistics company every day and analyzing and checking whether there are the following situations:

[0026] A. Decreased gross profit margin: The logistics company’s cumulative gross profit margin for several consecutive days has dropped by more than a certain threshold compared with the cumulative gross profit margin of the past 12 months or the original assessed gross profit margin;

[0027] B. Business volume shrinkage: The average transaction volume of the logistics company for several consecutive days is lower than a certain threshold compared with the average daily transaction volume in the previous 12 months; the average transaction volume for several consecutive days is lower than a certain threshold compared with the average daily transaction volume in the previous 12 months.

[0028] Preferably, an incentive method provided by the method is to set a part of the interest as a variable interest rate, and the variable interest rate is related to the risk factor. Assuming the variable interest rate is p%, and calculating the final variable interest rate f based on the risk factor K, the formula is:

[0029] f = min{1, K*m)*p%};

[0030] The variable interest that the user needs to pay is: advance amount*f*advance duration; if there is no abnormality during the process, the variable interest is 0.

[0031] Preferably, the abnormal records corresponding to the logistics company are processed according to the type of abnormality:

[0032] False business exceptions directly trigger the project fuse, restricting the issuance of orders and advance payments. Immediately conduct an exception investigation to find the cause. If the logistics company’s business is indeed false, the project will be terminated and the logistics company will be required to repay all the money within a time limit. If the business is not false, the current abnormal problem will be handled and rectified before the project continues;

[0033] Serious fund anomalies will not trigger the circuit breaker program, but will be notified to the management personnel, who will conduct an anomaly investigation within the specified time. If the logistics company does have fund risks, the platform will stop paying for freight. If it is not a fund issue, the logistics company will be required to handle and rectify the current anomaly.

[0034] After recording common operational anomalies, the management staff will be notified to investigate. If the situation persists and the operating conditions cannot be improved, the platform must stop advance payment of freight.

[0035] Preferably, the types of exceptions are divided into levels from high to low according to their severity: business false exceptions are graded as major exceptions, financial exceptions are graded as severe exceptions, and operating exceptions are graded as ordinary exceptions; when pushing exception messages, different handlers are designated according to the exception type, and they can be reported step by step according to the type level.

[0036] The present invention has the following advantages:

[0037] 1. The present invention classifies the abnormal situations that occur, and identifies the two types of abnormal situations, namely, business false abnormalities and capital abnormalities, through corresponding models and algorithms, so as to timely discover abnormalities that will cause major or serious risks to the platform's advance payment of freight, and promptly notify relevant personnel to conduct investigations and processing, thereby reducing the situation where abnormalities cause platform losses. On the other hand, based on past abnormal records, the risk coefficient is calculated, and incentives are set based on the risk coefficient to reward and punish logistics companies. On the one hand, the logistics companies are encouraged to operate and operate the system in accordance with the specifications of the network freight platform, and on the other hand, the platform abnormal management costs and business operation abnormal risks are reduced. The latter is to encourage logistics companies to execute the logistics process and operate the system of this platform in accordance with the specifications, so as to avoid the misleading judgment of abnormal categories caused by these situations and reduce the difficulty and time of management personnel to investigate.

[0038] 2. The present invention promptly and accurately identifies and judges the most risky business false anomalies. On the one hand, it converts multiple factors involved in judging whether a business is false into indicator variables, and then constructs a logistic regression model and trains to obtain the final judgment model, so as to accurately judge whether there are business false anomalies based on various factors, making the judgment result more reliable. On the other hand, since the actual vehicle trajectory is one of the important data sources for analyzing the factors involved, the present invention also uses an abnormal point identification method to identify and eliminate abnormal points in the collected vehicle trajectory data, so as to obtain a vehicle trajectory that is more in line with the actual situation and improve the accuracy of the judgment result.

[0039] 3. The present invention also takes into account the potential risk of advance payment due to operational difficulties when logistics companies operate in accordance with the platform's regulatory requirements. Therefore, by setting classification definitions and active monitoring of operational anomalies, even if various situations belonging to operational anomalies are found, logistics companies with operational difficulties can be identified, thereby avoiding the risk of advance payment caused by such logistics companies through corresponding exception handling methods.

[0040] 4. The present invention also divides the risk severity level according to the risk of the situation corresponding to the exception type, so as to specify different handlers according to the exception type when pushing exception messages, and can report step by step according to the type level, thereby avoiding the problem of exception reporting and handling without distinguishing the priorities and confusing the order due to the lack of level division. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 The present invention is a flowchart of a method for abnormal early warning control of network freight advance payment.

[0042] Figure 2 The present invention is a flow chart for performing freight advance payment and advance payment exception monitoring in a platform using the present invention. DETAILED DESCRIPTION

[0043] The specific implementation modes of the present invention are further explained in detail below by describing the embodiments with reference to the accompanying drawings, so as to help those skilled in the art to have a more complete, accurate and in-depth understanding of the inventive concept and technical solution of the present invention.

[0044] like Figure 1 and Figure 2 As shown, the process of freight advance payment and advance payment exception monitoring in the platform includes: project credit configuration, exception warning definition and monitoring execution, logistics company self-payment, platform advance payment and post-anomaly processing.

[0045] Generally speaking, the logistics platform conducts credit assessment and credit management for logistics companies in the form of projects. The projects correspond to logistics companies, customers (Party A enterprises), transportation routes, cargo types, prices, and payment cycles. According to the scale of the project's transportation business, the credit status, operating conditions, and asset status of the customer, and the credit status, operating conditions, and asset status of the logistics company, the project is evaluated for credit, and the total project funding demand T, the user's investment ratio P, and the platform's advance payment ratio 1-P are determined. The user completes the self-payment first, and the user's self-payment amount M1 = T*P, and the platform's advance payment amount M2 = T*(1-P). The logistics company publishes the waybill on the platform and pays the freight. At the beginning, the logistics company must first use its own funds to complete the self-payment, and the total self-payment amount is M1. Only when the user's self-payment amount exceeds M1 can the platform's funds be used for advance payment. After the user completes the self-payment amount M1 on the platform, the platform's funds can be used for subsequent freight advance payment, and the maximum amount that can be used is M2. After the user repays the previous loan, the platform funds can be used in a rolling manner. That is, if the user can currently use the platform to advance the amount M3, the advance amount P1, and the repaid amount R1, then M3=M2-P1+R1.

[0046] In the above advance payment process, the logistics company may use the process to withdraw funds from the platform or fail to repay due to the company's financial constraints, which will bring significant losses to the online freight platform. To solve these problems, the present invention discloses a method for abnormal early warning control of online freight freight advance payment.

[0047] When this method is implemented, first, in the project abnormality monitoring configuration stage, the classification definition of the abnormality type is completed, a judgment model for judging business false type abnormalities is constructed, and the corresponding algorithm rules related to abnormality monitoring are set. In the advance payment process, the risk coefficient of each logistics company is calculated according to the abnormality type, and the logistics company is incentivized according to the incentive measures determined thereby. After that, the project abnormality monitoring is executed, and a scheduled task is set. It is executed once every day at 9 am and 3 pm. The monitoring is performed according to the abnormality definition and abnormality algorithm rules, and abnormalities are actively discovered. Once an abnormality is found, a message is immediately pushed to the administrator. The abnormal message supports SMS, WeChat, and PC-side in-station messages. Multiple abnormality handlers can be maintained for the same abnormality type, and different handlers can be specified according to the abnormality type to realize the reporting of the abnormal message level by level according to the type level. The level of business false type abnormality is major abnormality, the level of capital type abnormality is serious abnormality, and the level of operation type abnormality is ordinary abnormality. Then, corresponding processing is performed according to the abnormality type.

[0048] This method also specifically includes the following contents:

[0049] 1. Abnormal classification: Classify the freight advance payment abnormalities on the platform. The abnormality types include false business abnormalities, capital abnormalities and operating abnormalities.

[0050] In the above-mentioned logistics process and fund transaction process, the data collected by the platform can reflect various anomalies that may cause advance payment risks. According to the causes of the anomalies, the anomalies reflected in the data can be divided into: business fraud anomalies, fund anomalies and operation anomalies.

[0051] False business anomalies: refers to the abnormal situation of cashing out through the platform through the fabricated business of false waybills. This type of anomaly is the most serious anomaly, and the actual logistics data is inconsistent with the content of the waybill.

[0052] Fund anomalies: refers to serious fund arrears problems in financial transaction data such as freight and repayment, which are serious anomalies indicating that the logistics company has fund anomalies and insufficient repayment ability.

[0053] Operational anomalies: refers to the situation where, while the logistics process and capital flow are completed normally, the logistics company's operating conditions continue to be poor for a period of time, resulting in a significantly higher risk of freight advance payment in the future. Since there are no anomalies in the process, the risk is a prediction of the logistics company's poor operation and reduced repayment ability, and therefore it is a common anomaly.

[0054] 2. Determination of business false anomalies: Build a judgment model for judging business false anomalies, obtain indicator data related to the model, and use the judgment model to determine whether there are business false anomalies and record them.

[0055] The factors involved in judging whether a business is fake come from the data collected during the actual logistics process and the data provided by the waybill, including vehicle tracks, waybill routes, the identity of the operator of the corresponding APP on the operating platform, receipt records, transportation prices, vehicle locations, pick-up and unloading locations and times in the waybill, and actual payment information, etc. The indicators, corresponding meanings, and values ​​required to determine whether a waybill is fake are shown in Table 1.

[0056] Table 1: Indicator variables, meanings and values

[0057] <![CDATA[ variable ]]> <![CDATA[ significance ]]> <![CDATA[ Assignment ]]> X1 The vehicle track does not match the waybill route 1 = yes, 2 = no X2 The APP is operated by someone other than the driver 1 = yes, 2 = no X3 There is no consignee signature record on the waybill 1 = yes, 2 = no X4 The shipping price is higher than the mean by a certain threshold (30%) 1 = yes, 2 = no X5 The vehicle does not appear at the pick-up and unloading place at the specified time 1 = yes, 2 = no X6 The driver does not collect the freight 1 = yes, 2 = no

[0058] A logistic regression model is established, and the model formula is as follows:

[0059]

[0060] Among them, P is the probability, β i(i=0,1,2,...,6) represents the regression coefficient, which is the coefficient to be determined, and Xi(i=0,1,2,...,6) is the indicator variable. The loss function is defined as L(a,y)=-ylog(a)-(1-y)log(1-a), where y represents the true label and a represents the predicted result. The indicator is entered into the Logistics model (i.e., the logistic regression model) according to the backward stepwise regression, and each regression coefficient is determined to finally obtain a determined judgment model.

[0061] For a waybill, the collected data is processed and the indicators are extracted and brought into the above judgment model to calculate the P value. A critical value is set, such as 0.5. When P is greater than the critical value, it indicates that the waybill is false; when P is not greater than the critical value, it is considered to be true. For waybills judged to be false, they are recorded as business false exceptions, and an exception message is generated and pushed to relevant management personnel.

[0062] In order to judge whether the waybill is false or not, it is necessary to accurately determine the actual vehicle trajectory in the logistics process, so as to identify the situation where the vehicle does not appear at the pickup place at the specified time and the vehicle trajectory does not match the waybill route. The drawing of the vehicle trajectory first requires the collection of vehicle coordinates. There are generally three ways to collect vehicle coordinates: app positioning, mobile phone base station positioning, and Beidou positioning. The collected coordinates and actual positions may sometimes deviate. In order to more accurately judge whether there is an abnormality in the vehicle trajectory, this paper constructs a distance-based outlier point recognition method to quickly identify abnormal coordinates and correct the vehicle trajectory.

[0063] (1) First, the driving trajectory of the transport vehicle is segmented. The time interval threshold is set to half an hour. Starting from the first coordinate obtained from the waybill, the trajectory drawn from the coordinates obtained within the first half an hour is the first segment, and the trajectory drawn from the coordinates obtained within the second half an hour is the second segment. This process is repeated in this way until all segments are obtained.

[0064] (2) Identify outliers in each segment, including the following specific steps.

[0065] a) Calculate the distance between each point (lonB, latB) in the segment and the starting point (lonA, latA)

[0066]

[0067] Where R is the radius of the Earth.

[0068] b) The actual distance traveled is dist2 = s*t, where s is the speed of the vehicle at that point and t is the time difference from that point to the starting point.

[0069] c) If If the value of is greater than the set comparison threshold, such as 5%, the point is considered to be an abnormal point.

[0070] (3) Delete the abnormal points in each segment and redraw the trajectory. The redrawn vehicle trajectory will be more consistent with the actual vehicle operation route.

[0071] 3. Determination of fund anomalies: Determine whether there are fund anomalies based on fund transaction data and record them.

[0072] Obtain the financial transaction data of each logistics company every day and check whether there are the following situations:

[0073] A. Delayed payment of freight: The logistics company delays paying the driver's freight for more than the set number of days, and the amount exceeds the set amount threshold. The set number of days threshold and the set amount threshold are configured by the management personnel, such as 30 days and RMB 30,000 respectively, which are defined specifically according to the business situation.

[0074] B. Failure to repay on time: The logistics company fails to repay the loan within the specified time.

[0075] The above situation indicates that the logistics company may have abnormal funds and a significant decline in repayment ability, which is classified as a serious abnormality. A record of a logistics company's abnormal funds occurs, an abnormal message is generated and pushed to relevant managers.

[0076] The above-mentioned contents 2 and 3 can be executed in any order and can be executed simultaneously.

[0077] 4. Calculation of risk factor: Conduct risk assessment based on recorded data of false business anomalies and financial anomalies to determine the risk factor.

[0078] In order to encourage logistics companies to implement normal logistics processes according to platform regulations and pay freight and repayments in full and on time, and to reduce the risk of logistics companies making false business judgments and insufficient funds due to operational errors and failure to implement logistics processes according to regulations, this method also sets a risk coefficient, and encourages logistics companies that comply with platform regulations based on the risk coefficient. The algorithm for the risk coefficient K is as follows:

[0079]

[0080] In the above formula, the total number of waybills is n. The weight of the business false class anomaly is w1 (between 0-100%). If any business false class anomaly occurs in a waybill, the business false class anomaly value t i The weight of the abnormal funds category is w2 (between 0 and 100%), w1+w2=100%. If any abnormal funds category appears in a shipping order, the abnormal funds category value s i It is recorded as 1, otherwise it is recorded as 0.

[0081] The above-mentioned risk factors are applied to incentives in order to distinguish logistics companies with normal business and repayment ability from those that set up false business or have no repayment ability. Incentives can reduce the former's abnormal identification errors caused by non-standard execution of logistics processes or improper operations, thereby reducing the difficulty and time for managers to conduct investigations.

[0082] 5. Determination of operational anomalies: Determine whether there are any operational anomalies based on the logistics company’s operating data on the platform and record them.

[0083] Obtain the financial transaction data of each logistics company every day and analyze and check whether there are the following situations:

[0084] A. Decreased gross profit margin: The logistics company’s cumulative gross profit margin for several consecutive days drops by more than a certain threshold, such as 4%, compared with the cumulative gross profit margin of the past 12 months or the original assessed gross profit margin.

[0085] B. Business volume shrinkage: The average transaction volume of the logistics company for several consecutive days is lower than a certain threshold value, such as 50%, compared with the average daily transaction volume in the previous 12 months; the average transaction volume for several consecutive days is lower than a certain threshold value, such as 30%.

[0086] The above situation can be judged as poor management of the logistics company and reduced repayment ability, which is an operational anomaly and is classified as a general anomaly. The number of days for the above situation to last is set by the management personnel based on experience or historical data. After the above situation occurs, the platform records it as an operational anomaly, generates an abnormal message and pushes it to the relevant management personnel, indicating that the freight advance payment needs to be stopped.

[0087] The above contents and contents 2 to 4 can be executed in any order and can be executed simultaneously.

[0088] 6. Exception handling: Based on the corresponding exception records of the logistics company, handle them according to the type of exception and risk factor.

[0089] For major anomalies such as business fraud, the project will be directly triggered to fuse, restricting the issuance of orders and advance payments. The cause will be immediately found. If the logistics company's business is indeed fraudulent and it is cashing out through advance payments on the platform, the project will be terminated and the logistics company will be required to repay the full amount within a time limit. If it is other special circumstances, and the business is not fraudulent, the logistics company will be required to make rectifications, and the project will continue after the current abnormal problem is resolved and rectified;

[0090] In the case of serious financial anomalies, the circuit breaker program will not be triggered, but the management personnel will be notified to conduct an abnormal investigation within the specified time. If the logistics company does have financial risks and the situation persists, the platform will stop advancing the freight; if it is other special circumstances, not financial issues, the logistics company will be required to complete the rectification of the current abnormal problems, that is, pay all overdue payments.

[0091] For common operational anomalies, if the situation persists and the operating conditions cannot be improved, the platform must stop advance payment of freight.

[0092] In order to encourage logistics companies to operate and manage the system in accordance with the standard requirements of the online freight platform, and at the same time reduce the platform's abnormal management costs and the risk of abnormal business operations, this method provides an incentive method by taking out a portion of the interest and setting it as a variable interest rate. The variable interest rate is related to the risk factor. The exceptions here are major business falsehood exceptions and serious financial exceptions, and they have been verified to be failures to operate or manage the system in accordance with the standards, and are not real business falsehoods or financial exceptions.

[0093] Assume the variable interest rate is p%, and calculate the final variable interest rate f based on the risk factor K. The formula is:

[0094] f = min{1, K*m)*p%};

[0095] The variable interest that users need to pay is: advance amount * f * advance duration. If there is no abnormality during the process, the variable interest is 0.

[0096] The above incentive method based on the risk factor K takes into account the severity (i.e., the corresponding weight) and the number of abnormalities of different types, and can therefore effectively encourage companies to follow the logistics process and make timely payments. In addition, other incentive methods can also be used as an alternative to the risk factor K, such as calculating additional deposits based on the risk factor K or fines after the above-mentioned abnormalities occur.

[0097] The present invention is described above by way of example in conjunction with the accompanying drawings. It is obvious that the specific implementation of the present invention is not limited to the above-mentioned method. As long as various non-substantial improvements are made using the inventive concept and technical solution of the present invention, or the inventive concept and technical solution are directly applied to other occasions without improvement, they are all within the protection scope of the present invention.

Claims

1. A method for abnormal early warning control of network freight advance payment, characterized by: include: First, in the project abnormality monitoring configuration stage, the classification definition of abnormality types is completed, a judgment model for judging business false abnormalities is constructed, and corresponding algorithm rules related to abnormality monitoring are set. In the advance payment process, the risk coefficient is calculated based on the abnormal records of the corresponding logistics company, and the logistics company is incentivized according to the determined incentive measures; then, the project abnormality monitoring is executed, and monitoring is performed regularly according to the abnormality definition and abnormality algorithm rules to actively discover abnormalities; Once an abnormality is found, a message will be immediately pushed to the relevant personnel after being recorded; then corresponding processing will be carried out according to the type of abnormality; The types of anomalies include at least business false anomalies and capital anomalies. Business false anomalies refer to the discrepancy between the actual logistics data and the waybill contents. Capital anomalies refer to the arrears of funds in the data of capital transactions such as freight and repayment. The algorithm of risk coefficient K is as follows: In the above formula, the total number of waybills is n, the weight of the business false class anomaly is w1, and as long as any business false class anomaly occurs in a waybill, the business false class anomaly value t i is recorded as 1, otherwise it is recorded as 0; the weight of the fund abnormality class is w2. If any fund abnormality occurs in a waybill, the fund abnormality value s i W1 is recorded as 1, otherwise it is recorded as 0; w1 and w2 are both between 0-100%, w1+w2=100%; The judgment of business false anomalies includes: the factors involved in judging whether the business is false are derived from the data collected in the actual logistics process and the data provided by the waybill. The collected data is processed and the indicators are extracted and substituted into the judgment model to calculate the probability value. The calculated result is compared with the critical value to judge whether the business corresponding to the waybill is false; the judgment model is established based on the logistic regression model as follows: Among them, Xi (i = 0, 1, 2, ..., 6) represents the indicator variable, X1-X6 represent respectively: the vehicle trajectory does not match the waybill route, the driver does not operate the APP, the waybill does not have the consignee's signature record, the transportation price is higher than the mean by a certain threshold, the vehicle does not appear at the pick-up and unloading place at the specified time, and the driver does not collect the freight himself; P is the probability, β ； (i=0,1,2,…,6) represents the regression coefficient, which is the coefficient to be determined; the loss function is defined as L(a,y)=-ylog(a)-(1-y)log(1-a), y represents the true label, and a represents the predicted result; the indicator is entered into the logistic regression model according to the backward stepwise regression, and each regression coefficient is determined, and finally a determined judgment model is obtained; The determination of fund anomalies includes: obtaining the fund transaction data of each logistics company every day and checking whether there are the following situations: A. Delayed payment of freight: The logistics company delays paying the driver freight for more than the set number of days, and the amount exceeds the set amount threshold; B. Failure to repay on time: The logistics company fails to repay the loan within the specified time.

2. A method for abnormal early warning control of network freight advance payment according to claim 1, characterized in that: Actual logistics data includes vehicle trajectories, which are used to identify situations where a vehicle does not appear at the pickup location at the specified time and where the vehicle trajectory does not match the waybill route; This method uses a distance-based outlier recognition model to quickly identify abnormal coordinates and thus correct the vehicle trajectory; The outlier identification method comprises: (1) First, the driving trajectory of the transport vehicle is segmented; (2) Identify outliers in each segment; (3) Delete the outliers in each segment and redraw the trajectory.

3. A method for abnormal early warning control of network freight advance payment according to claim 2, characterized in that: The step (2) comprises the following specific steps: a) Calculate the distance between each point (lonB, latB) in the segment and the starting point (lonA, latA) Where R is the radius of the earth; b) The actual distance traveled is dist2 = s*t, where s is the speed of the vehicle at that point and t is the time difference from that point to the starting point; c) If If the value of is greater than the set comparison threshold, the point is considered to be an abnormal point.

4. A method for abnormal early warning control of network freight advance payment according to claim 3, characterized in that: The abnormality types also include operational abnormalities, which refer to the poor operating conditions of a logistics company over a period of time. The determination of operational abnormalities includes: obtaining the capital transaction data of each logistics company every day and analyzing and checking whether there are the following situations: A. Decreased gross profit margin: The logistics company’s cumulative gross profit margin for several consecutive days has dropped by more than a certain threshold compared with the cumulative gross profit margin of the past 12 months or the original assessed gross profit margin; B. Business volume shrinkage: The average transaction volume of a logistics company for several consecutive days is lower than a certain threshold compared with the average daily transaction volume in the previous 12 months.

5. A method for abnormal early warning control of network freight advance payment according to any one of claims 1 to 3, characterized in that: One incentive provided by this method is to set a portion of the interest as a variable interest rate. The variable interest rate is related to the risk factor. Suppose the variable interest rate is p%, and the final variable interest rate f is calculated based on the risk factor K. The formula is: f = min{1, K*m)*p%}; The variable interest that the user needs to pay is: advance amount*f*advance duration; if there is no abnormality during the process, the variable interest is 0.

6. A method for abnormal early warning control of network freight advance payment according to claim 4, characterized in that: Based on the abnormal records corresponding to the logistics company, handle them according to the type of abnormality: False business exceptions directly trigger the project fuse, restricting the issuance of orders and advance payments. Immediately conduct an exception investigation to find the cause. If the logistics company’s business is indeed false, the project will be terminated and the logistics company will be required to repay all the money within a time limit. If the business is not false, the current abnormal problem will be handled and rectified before the project continues; Serious fund anomalies will not trigger the circuit breaker program, but will be notified to the management staff, who will conduct an anomaly investigation within the specified time. If the logistics company does have fund risks, the platform will stop paying the freight in advance. If it is not a funding issue, the logistics company is required to complete the rectification of the current abnormal problem; After recording common operational anomalies, the management staff will be notified to investigate. If the situation persists and the operating conditions cannot be improved, the platform must stop advance payment of freight.

7. A method for abnormal early warning control of network freight advance payment according to claim 6, characterized in that: The types of exceptions are divided into levels from high to low according to their severity: business false exceptions are classified as major exceptions, capital exceptions are classified as severe exceptions, and operating exceptions are classified as ordinary exceptions. When pushing exception messages, different handlers are designated according to the exception type, and they can be reported step by step according to the type level.

Images