Method and system for predicting bad debt risk of passenger payment behavior based on SaaS platform

Abstract

The invention discloses a bad debt risk prediction method and system based on SaaS platform passenger payment behaviors. The invention relates to the technical field of online car hailing. Introducing a D-S evidence theory algorithm, and setting a risk proposition and a risk-free proposition; based on the data set D, decomposing into an evidence set A and an evidence set B corresponding to a risk proposition and a risk-free proposition; endowing basic probability assignment (BPA) to the risk proposition and the risk-free proposition based on a bell-shaped function; calculating a risk factor E by applying a merging principle of a D-S evidence theory algorithm and combining the retrieved passenger information; by integrating multi-dimensional data such as real-time ordering information, historical payment records, credit scores, external blacklists and fraudulent behavior records of passengers, a more comprehensive and accurate passenger credit evaluation system can be constructed. And the bad debt probability of the order can be accurately predicted when the passenger places the order, so that measures can be taken in time for prevention.

Description

Technical Field

[0001] The present invention relates to the field of online car-hailing technology, specifically to platform payment, and in particular to a bad debt risk prediction method and system based on passenger payment behavior on a SaaS platform. Background Art

[0002] Designated driver travel services, as a convenient and necessary transportation solution, are mainly aimed at vehicle owners who are unable to drive their cars themselves due to specific situations (such as after drinking). This service provides car owners with a safe and reliable designated driver experience by deploying professional drivers, thereby ensuring road safety and personal well-being. However, the particularity of this service model leads to a relatively low repurchase cycle compared to other travel service areas. The root cause of this phenomenon is that the demand for designated drivers is often closely related to occasional events, rather than daily commuting or fixed travel needs, which in turn affects the user's continued frequency of use.

[0003] It is particularly noteworthy that the problem of bad debts of passengers in designated driver services has become increasingly prominent and has become one of the key factors restricting the healthy development of the industry. Bad debts refer to situations where passengers fail to pay service fees on time. There is still no clear boundary demarcation in the current designated driver market. Nevertheless, through in-depth analysis of existing literature, media reports and market observations, it is not difficult to find that there are significant differences in passengers' repurchase behavior and payment habits on different designated driver platforms, which provides clues for further exploring the causes of bad debts.

[0004] In order to deal with the bad debt problem, existing designated driver platforms generally adopt a variety of collection strategies including SMS reminders and phone calls. Although these measures can encourage some passengers to complete payment to a certain extent, their inherent lag and inefficiency not only make it difficult to identify potential risk passengers in real time and implement early intervention, but also may cause unnecessary interference to the daily lives of the majority of passengers who fulfill their contracts, affecting the user experience and the reputation of the platform.

[0005] In view of this, the designated driver industry urgently needs a more forward-looking and intelligent risk management mechanism. To this end, the present invention proposes a bad debt risk prediction method and system based on the payment behavior of passengers on the SaaS platform. Summary of the invention

[0006] In view of this, the present invention hopes to provide a bad debt risk prediction method and system based on the payment behavior of passengers on the SaaS platform to solve or alleviate the technical problems existing in the prior art, that is, how to predict the bad debt probability of the order when the passenger places an order, and take timely measures to prevent it, and at least provide a beneficial option for this; the technical solution of the present invention is implemented as follows:

[0007] First, the bad debt risk prediction method based on the payment behavior of passengers on the SaaS platform:

[0008] 1. Overview:

[0009] The present invention aims to predict and prevent bad debt problems in designated driver services. A comprehensive passenger credit assessment system is constructed by integrating the personal information submitted by passengers when placing orders, the risk control database of the designated driver service system, and external open data sources. Using advanced DS evidence theory algorithms, comprehensive analysis and risk assessment of multi-dimensional information such as passengers' payment behavior, credit history, blacklist records, and fraudulent behavior are performed to calculate risk factors to quantify the probability of bad debts. According to the results of risk assessment, corresponding preventive measures are taken in a timely manner, such as requiring high-risk passengers to pay outstanding fees first or restricting services to reduce the incidence of bad debts. At the same time, through detailed payment situation retrieval and external data source collision, the credit status of passengers is further verified to ensure the accuracy and reliability of risk assessment.

[0010] (II) Technical solution:

[0011] To achieve the above technical goals, when an activation instruction is received from the SaaS designated driver service platform, indicating that there are new passenger orders that require bad debt prediction and prevention processing, perform the following steps.

[0012] 2.1 Step S1, data reception and preprocessing:

[0013] Receive the personal information submitted by the passenger when placing an order, access the risk control database of the designated driver service system, retrieve the passenger's historical payment record, credit score, blacklist record and fraud record, and form data set D;

[0014] At the same time, the data set D is preprocessed, including data cleaning (removing invalid or erroneous data), and then normalized so that each element in the data set D has the same scalar, where the data set D = [a, b, c, d];

[0015] Among them, a is the normalized historical payment record sequence of the passenger, b is the normalized credit score, c is the normalized blacklist record ratio, and d is the normalized fraud record ratio.

[0016] 2.2 Step S2, risk assessment and calculation:

[0017] The DS evidence theory algorithm is introduced, and two propositions, risky and risk-free, are set; the merging principle of the DS evidence theory algorithm is applied, combined with the retrieved passenger information, to calculate the risk factor E.

[0018] 2.2.1 Step S200, setting the topic:

[0019] Set up two propositions, namely the risky proposition H1 and the risk-free proposition H2.

[0020] 2.2.2 Step S201, decomposing the data set D:

[0021] Based on the data set D, it is decomposed into two evidence sets A and B corresponding to the risky proposition H1 and the risk-free proposition H2;

[0022] S2010, threshold setting: For each selected feature of the data set D, set one or more thresholds to distinguish between risky and non-risky situations.

[0023] For example: Let fi(x) represent the i-th feature x in the data set D i A threshold function that returns 0 (no support) or 1 (support) depending on the feature value.

[0024] S2011, data partitioning: Based on the set threshold, each sample in the dataset D is divided into evidence supporting the risky proposition H1 or the risk-free proposition H2 by creating a binary vector; 1 indicates support for the proposition and 0 indicates non-support.

[0025] For example: For a total of n features x in the data set D i , calculate its corresponding evidence vector:

[0026] e=[f1(x1),f2(x2),…,fn(x n )].

[0027] According to the value in the evidence vector e, sample d is divided into evidence set A or B.

[0028] S2012, form an evidence set: all samples supporting the risky proposition H1 form evidence set A, and all samples supporting the risk-free proposition H2 form evidence set B.

[0029] For example, based on the value in the evidence vector e, sample d is divided into evidence set A or B.

[0030] 2.2.3 Step S202, assigning basic probability assignment BPA:

[0031] Use the bell-shaped function to assign basic probability assignments BPA to the risky proposition H1 and the risk-free proposition H2, denoted as mH1(A) and mH2(B) respectively. The bell-shaped function can calculate the corresponding basic probability assignment BPA based on the element values ​​in the data set D:

[0032]

[0033] Among them, BPA (x i ) indicates that in feature x iBPA value; x is a feature value in the data set D. μ is the center point of the bell-shaped function, indicating the most likely value or the most trusted point; α is the width parameter of the bell-shaped function, which controls the steepness of the function. The larger α is, the steeper the function is; the smaller α is, the flatter the function is.

[0034] 2.2.4 Step S203, applying the merging principle of the DS evidence theory algorithm:

[0035] According to the merging principle (Dempster-Shafer synthesis rule) of the DS evidence theory algorithm, the retrieved passenger information (i.e., evidence sets A and B) is combined to merge the evidence. The probability assignment after merging is expressed as m12(E), where F is the set of all possible propositions on the recognition frame:

[0036]

[0037] Among them, K is the conflict coefficient, which indicates the degree of inconsistency between evidences;

[0038] m1(A i ) and m2(B j ) are the basic probability assignments BPA of the corresponding propositions in the evidence sets A and B respectively;

[0039] A i and B j are subsets of evidence sets A and B respectively; F is the set of propositions on the identification framework.

[0040] 2.2.5 Step S204, calculate the risk factor E:

[0041] According to the combined probability assignment m12(E), the risk factor E is calculated. The risk factor E can be quantified as the ratio of the probability of the risky proposition H1 to the probability of the risk-free proposition H2:

[0042] 2.3 Step S3, risk assessment and preventive measures:

[0043] If the risk factor E exceeds the preset threshold T, and the passenger ID has unpaid orders in the past, a payment reminder will be sent to the passenger, requiring him to pay the outstanding fees first.

[0044] If the risk factor E exceeds the preset threshold T, but the passenger ID has no unpaid orders, or the risk factor E does not exceed the preset threshold T, the next step is to retrieve the detailed payment information.

[0045] 2.4 Step S4, detailed payment information retrieval:

[0046] Retrieve the passenger’s past payment history, including whether the payment has timed out; at the same time, perform a database collision with external black market data sources to check whether the passenger is involved in fraudulent behavior.

[0047] 2.4.1 Step S400, payment behavior analysis:

[0048] Traverse the passenger's payment records and mark all payment timeouts.

[0049] 2.4.2 Step S401, payment habit assessment:

[0050] Collect statistics on passengers’ payment frequency, payment amount distribution and common payment channels; assess whether there are fluctuations or inconsistencies in passengers’ payment habits.

[0051] 2.4.3 Step S402: External Black Market Data Source Collision

[0052] Compare the passenger’s payment records with fraud records in external black market data sources; check whether the passenger’s payment account, payment channel, and payment IP address appear in the fraud database;

[0053] Once a passenger is found to be involved in fraudulent behavior, risk identification will be carried out.

[0054] 2.4.4 Step S403, result summary and report

[0055] Based on the analysis results, a detailed payment retrieval report is generated, which includes payment timeouts, payment habit assessment results and fraud detection results information.

[0056] 2.5 Step S5, risk assessment result output:

[0057] Generate a risk assessment report based on the results of risk assessment and detailed payment information retrieval, including passengers' bad debt warning information and existing risk points.

[0058] (III) Mechanism for solving technical problems:

[0059] 3.1 Risk assessment model construction:

[0060] The DS evidence theory algorithm is used to process uncertain information. The algorithm can integrate evidence from multiple sources to assess the bad debt risk of passengers. Two propositions, risky and risk-free, are set, and the data set D (including the passenger's historical payment records, credit scores, etc.) is decomposed into multiple evidence sets.

[0061] Each set of evidence is assigned a basic probability assignment (BPA), which reflects the degree of support for the proposition. The assignment of BPA is based on a bell-shaped function, which ensures the rationality and accuracy of the assignment.

[0062] Applying the merging principle of the DS evidence theory algorithm, the information of all evidence sets is combined to calculate the risk factor E. The risk factor E is a quantitative indicator used to represent the bad debt probability of passenger orders.

[0063] 3.2 Risk assessment and preventive measures:

[0064] A preset risk threshold T is set to determine whether the bad debt risk of the passenger's order has reached a level that requires measures to be taken. At the same time, a more detailed search of the passenger's past payment records is performed, including whether the payment has timed out, whether the order amount is abnormal, whether there is abnormal behavior during the trip, etc. At the same time, it is collided with external black market data sources to check whether the passenger is involved in fraudulent behavior.

[0065] Second, the bad debt risk prediction system based on the payment behavior of passengers on the SaaS platform:

[0066] The system is used to implement the bad debt risk prediction method and system based on the SaaS platform passenger payment behavior mentioned above, which includes:

[0067] (1) Receive the personal information submitted by the passenger when placing an order, and access the data receiving and preprocessing module of the risk control database of the designated driver service system: retrieve the passenger's historical payment record, credit score, blacklist record, and fraud record. This constitutes data set D, which provides basic data for subsequent risk assessment.

[0068] This module is connected to the front-end interface of the designated driver service system to receive the order information of the passenger. At the same time, it interacts with the risk control database to obtain the historical data of the passenger. The processed data set D is passed to the risk assessment and calculation module.

[0069] (2) Introduce the DS evidence theory algorithm to calculate the risk factor E. Risk assessment and calculation module: Introduce the DS evidence theory algorithm and set two propositions: risky and risk-free. Decompose the data set D into two evidence sets A and B corresponding to the risky and risk-free propositions. Use the bell-shaped function to assign basic probability assignments BPA to the two propositions. Apply the merging principle of the DS evidence theory algorithm and combine the retrieved passenger information to calculate the risk factor E.

[0070] (3) A risk assessment and prevention module that compares the risk factor E with the preset threshold T to determine the bad debt risk of the passenger order.

[0071] (4) Payment retrieval module that retrieves the passenger’s past payment history: It also performs a database collision with external black market data sources to check whether the passenger is involved in fraudulent behavior.

[0072] (5) A risk assessment result output module that generates a risk assessment report based on the results of risk assessment and detailed payment information retrieval: the report includes the passenger's bad debt warning information and existing risk points.

[0073] Compared with the prior art, the present invention has the following beneficial effects:

[0074] 1. Improve the accuracy of bad debt prediction: By integrating multi-dimensional data such as passengers' real-time order information, historical payment records, credit scores, external blacklists and fraud records, the present invention can build a more comprehensive and accurate passenger credit evaluation system. This helps to accurately predict the bad debt probability of an order when the passenger places an order, so that timely measures can be taken to prevent it.

[0075] 2. Enhanced risk control capabilities: By introducing advanced risk assessment technologies such as the DS evidence theory algorithm, the present invention can process uncertain information and integrate evidence from multiple sources to assess the bad debt risk of passengers. This enables the designated driver service system to more intelligently identify potential high-risk passengers and take corresponding preventive measures, thereby reducing the incidence of bad debts.

[0076] 3. Improve operational efficiency: The present invention realizes the automation and intelligence of risk assessment, reduces manual intervention and subjective judgment, and improves the efficiency and accuracy of risk assessment. It helps the designated driver service system to respond to market changes more quickly, optimize resource allocation, and improve overall operational efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0078] Figure 1 It is a schematic diagram of the method composition of the present invention;

[0079] Figure 2 It is a schematic diagram of the system composition of the present invention;

[0080] Figure 3 The present invention is a schematic diagram of the method execution flow. DETAILED DESCRIPTION

[0081] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings. In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below;

[0082] It should be noted that the various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments, and the same or similar parts between the various embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description.

[0083] Explanation of relevant terms:

[0084] (1) Designated driver service system: an online platform that provides designated driver reservation, dispatch, payment and other services.

[0085] (2) Risk control database: A database that stores the credit information and transaction records of passengers and drivers and is used for risk control.

[0086] (3) Historical payment records: the passenger’s past payment status and historical transaction records in the designated driver service.

[0087] (4) Credit score: A credit rating or score calculated based on factors such as the passenger’s payment behavior and contract performance.

[0088] (5) Blacklist records: A list of passengers or drivers with bad behavior or violation records, used for risk warning and service restriction.

[0089] (6) Fraudulent behavior: dishonest behavior such as deception, false transactions, etc. adopted by passengers or drivers in designated driving services.

[0090] (7) DS Evidence Theory Algorithm: A reasoning algorithm used to process uncertain information and make decisions by combining evidence from different sources.

[0091] (8) Basic Probability Assignment (BPA): In DS evidence theory, the initial probability value assigned to each proposition or evidence is used for subsequent evidence merging and reasoning.

[0092] (9) External black market data sources: Credit information data sources such as blacklists and fraudulent activities from other platforms or institutions, which are used to enhance the accuracy of risk assessment.

[0093] Embodiment 1: Figure 1 , 3As shown, this embodiment discloses a bad debt risk prediction method based on the payment behavior of passengers on the SaaS platform; when the SaaS designated driver service platform receives an order from a passenger, it is necessary to predict the bad debt probability of the order when the passenger places the order, so as to take measures to prevent the bad debt risk in time. The specific execution process includes the following steps S1 to S5.

[0094] In this embodiment, regarding step S1, data reception and preprocessing: when a passenger places an order, the system receives personal information submitted by the passenger, including name, mobile phone number, ID number, etc. The system accesses the risk control database of the designated driver service system to retrieve the passenger's historical payment records, credit scores, blacklist records, and fraud records.

[0095] The received data is cleaned to remove invalid or erroneous data, such as missing values, outliers, etc. Normalization is performed so that each element in the data set D has the same scalar. Specifically, the passenger's historical payment record sequence, credit score, blacklist record ratio, and fraud record ratio are normalized to obtain the data set D = [a, b, c, d].

[0096] Among them, a is the normalized historical payment record sequence of the passenger, b is the normalized credit score, c is the normalized blacklist record ratio, and d is the normalized fraud record ratio.

[0097] In this embodiment, regarding step S2, risk assessment and calculation: introduce the DS evidence theory algorithm, set two propositions of risk and no risk; apply the merging principle of the DS evidence theory algorithm, combine the retrieved passenger information, and calculate the risk factor E.

[0098] Specifically, set propositions (step S200): set two propositions: a risky proposition H1 and a risk-free proposition H2.

[0099] Specifically, decompose the data set D (step S201):

[0100] Threshold setting (S2010): For each feature in the dataset D, set one or more thresholds to distinguish between risky and non-risky situations. Let fi(x) represent the i-th feature x in the dataset D. i A threshold function that returns 0 (no support) or 1 (support) depending on the feature value.

[0101] For example, for a historical payment record sequence, a threshold of payment timeouts exceeding a certain number of times may be set as risky; for a credit score, a threshold of a score below a certain score may be set as risky.

[0102] Specifically, data partitioning (S2011): according to a set threshold, each sample in the data set D is divided into evidence supporting the risky proposition H1 or the risk-free proposition H2.

[0103] For example, if the number of payment timeouts in the passenger's historical payment record exceeds a threshold, the sample is classified as evidence supporting the risky proposition H1. : According to the set threshold, a binary vector is created to classify each sample in the data set D as evidence supporting the risky proposition H1 or the risk-free proposition H2; 1 indicates support for the proposition, and 0 indicates non-support.

[0104] For example: For a total of n features x in the data set D i , calculate its corresponding evidence vector:

[0105] e=[f1(x1),f2(x2),…,fn(x n )].

[0106] Specifically, an evidence set is formed (S2012): all samples supporting the risky proposition H1 are formed into an evidence set A, and all samples supporting the risk-free proposition H2 are formed into an evidence set B.

[0107] Exemplarily, sample d is divided into evidence set A or B according to the value in evidence vector e.

[0108] Specifically, step S202, assigning basic probability assignment BPA: using a bell-shaped function to assign basic probability assignment BPA to the risky proposition H1 and the risk-free proposition H2, respectively denoted as mH1(A) and mH2(B). The bell-shaped function can calculate the corresponding basic probability assignment BPA according to the element value in the data set D:

[0109]

[0110] Among them, BPA (x i ) indicates that in feature x i BPA value; x is a feature value in the data set D. μ is the center point of the bell-shaped function, indicating the most likely value or the most trusted point; α is the width parameter of the bell-shaped function, which controls the steepness of the function. The larger α is, the steeper the function is; the smaller α is, the flatter the function is.

[0111] Specifically, in step S203, according to the merging principle (Dempster-Shafer synthesis rule) of the DS evidence theory algorithm, the retrieved passenger information (i.e., evidence sets A and B) is combined to merge the evidence. The probability assignment after the merger is expressed as m12(E), where F is the set of all possible propositions on the recognition framework:

[0112] Among them, K is the conflict coefficient, which indicates the degree of inconsistency between evidences;

[0113] m1(A i ) and m2(B j ) are the basic probability assignments BPA of the corresponding propositions in the evidence sets A and B respectively;

[0114] A i and B j are subsets of evidence sets A and B respectively; F is the set of propositions on the identification framework.

[0115] Specifically, step S204, calculate the risk factor E: calculate the risk factor E according to the combined probability assignment m12(E). The risk factor E can be quantified as the ratio of the probability of the risky proposition H1 to the probability of the risk-free proposition H2:

[0116] According to the merging principle of the DS evidence theory algorithm, the evidence is merged by combining the retrieved passenger information (i.e., evidence sets A and B). The probability assignment after merging is expressed as m12(E), where E is the set of all possible propositions on the recognition frame. The probability assignment after merging is obtained by calculating the conflict coefficient K and the basic probability assignment BPA of the corresponding proposition.

[0117] In this embodiment, regarding step S3, risk judgment and preventive measures: if the risk factor E exceeds the preset threshold T, and the passenger ID has unpaid orders in the past, a payment reminder is sent to the passenger, requiring the passenger to pay the outstanding fees first. If the risk factor E exceeds the preset threshold T, but the passenger ID has no unpaid orders, or the risk factor E does not exceed the preset threshold T, the next step of detailed payment status retrieval is entered. By sending payment reminders in a timely manner or requiring prepayment and other measures, the occurrence of bad debt risks can be prevented.

[0118] In this embodiment, regarding step S4, detailed payment information retrieval: retrieve the passenger's past payment information, including whether the payment has timed out; at the same time, perform a database collision with an external black market data source to check whether the passenger is involved in fraudulent behavior.

[0119] Specifically, step S400, payment behavior analysis: traverse the passenger's payment records and mark all payment timeouts. For example, check whether there are payment timeouts exceeding a certain number of times in the passenger's past payment records.

[0120] Specifically, step S401, payment habit assessment: statistics on the payment frequency, payment amount distribution and common payment channel information of passengers. For example, analyze whether the payment habits of passengers are stable and normal, and whether there are abnormal fluctuations or inconsistencies.

[0121] Specifically, in step S402, the passenger's payment record is compared with the fraud records in the external black market data source. For example, the passenger's payment account, payment channel and payment IP address are checked to see if they appear in the fraud database. Once the passenger is found to be involved in fraud, a risk identification is performed.

[0122] Specifically, result summary and report (step S403): The scores of the overall output of the model are classified into levels, such as high risk, medium risk, low risk, etc. According to the classification level, take corresponding preventive measures, such as prepayment of orders for medium and low risks, and restriction of orders for high risks. Calculate the bad debt risk score, consider characteristic items such as large orders, payment timeouts, non-payment across platforms, black market data sources, first orders from new customers, etc., and give corresponding scores. For example, the larger the amount of large orders, the higher the score, and the longer the payment timeout, the higher the score, etc.

[0123] In this embodiment, regarding step S5, the risk assessment result output: a risk assessment report is generated based on the results of the risk assessment and detailed payment situation retrieval. The report includes the passenger's bad debt warning information and existing risk points. For example, the report can clearly indicate the passenger's bad debt risk level, payment timeout, records involving fraudulent behavior, and other information. The risk assessment report is submitted to the relevant decision-making department or risk management team so that they can take further measures to prevent bad debt risks based on the information in the report.

[0124] It is understandable that through the implementation of this method, the SaaS driver service platform can predict the bad debt probability of the order when the passenger places the order, and take timely preventive measures. This helps to reduce the probability of bad debt risk and improve the platform's capital recovery rate and operational efficiency. This method combines multi-dimensional information such as the passenger's historical payment records, credit scores, blacklist records, and fraud records, and uses the DS evidence theory algorithm to perform risk assessment and calculation. By setting two propositions, risky and risk-free, and introducing a bell-shaped function to assign a basic probability assignment BPA, the risk factor E is finally calculated based on the combined probability assignment. According to the comparison results of the size of the risk factor E and the preset threshold, take corresponding preventive measures to reduce the risk of bad debts.

[0125] Embodiment 2: This embodiment further provides a Python execution program of the solution described in Embodiment 1:

[0126]

[0127]

[0128]

[0129]

[0130]

[0131] In the above program, the system receives the personal information submitted by the passenger and accesses the risk control database to retrieve relevant records. The received data is cleaned to remove invalid or erroneous data. Normalization is performed so that each feature has the same scalar value for subsequent calculations.

[0132] The DS evidence theory algorithm is introduced, and two propositions, risk and no risk, are set. A threshold is set to distinguish between risky and no risk situations, and the data is divided into evidence supporting different propositions according to the threshold. A bell-shaped function is used to assign basic probability assignments (BPA) to risky and no risk propositions.

[0133] According to the merging principle of DS evidence theory, the evidence is merged with the retrieved passenger information to obtain the probability assignment after the merger. The risk factor E is calculated and quantified as the ratio of the probability of a risky proposition to the probability of a risk-free proposition.

[0134] The risk level of the passenger is determined based on the risk factor E and the preset threshold T. If the risk factor E exceeds the preset threshold T and the passenger has unpaid orders, a payment reminder is sent. Otherwise, the next step is to retrieve the detailed payment information.

[0135] Retrieve the passenger's past payment history, including whether the payment has timed out, etc. Compare with external black market data sources to check whether the passenger is involved in fraudulent behavior. Based on the retrieval results, conduct payment behavior analysis, payment habit assessment, and fraud detection.

[0136] Based on the results of risk assessment and detailed payment information retrieval, a risk assessment report is generated. The report includes the passenger's bad debt warning information and existing risk points, and is submitted to the relevant decision-making department or risk management team.

[0137] All the above embodiments only express the implementation methods of the relevant practical applications of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be based on the attached claims.

[0138] For those skilled in the art, it can be further appreciated 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. Professional and technical personnel can 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.

[0139] At the same time, those skilled in the art can understand that all or part of the processes in all the above-mentioned embodiments can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided in this application and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

Claims

1. A bad debt risk prediction method based on the payment behavior of passengers on the SaaS platform, characterized by: After receiving the activation instruction sent by the SaaS designated driver service platform, start the following steps: S1, receives the personal information submitted by the passenger when placing an order, accesses the risk control database of the designated driver service system, retrieves the passenger's historical payment record, credit score, blacklist record and fraud record, and forms data set D; S2, introduce the DS evidence theory algorithm, set two propositions: risky and risk-free; apply the merging principle of the DS evidence theory algorithm, combine the retrieved passenger information, and calculate the risk factor E S3: If the risk factor E exceeds the preset threshold T, and the passenger ID has unpaid orders in the past, a payment reminder is sent to the passenger, requiring the passenger to pay the outstanding fees first; if the risk factor E exceeds the preset threshold T, but the passenger ID has no unpaid orders, or the risk factor E does not exceed the preset threshold T, then enter S4; S4, retrieves the passenger’s past payment history, including whether the payment has timed out; at the same time, performs a database collision with external black market data sources to check whether the passenger is involved in fraud.

2. The bad debt risk prediction method according to claim 1, characterized in that: In S1, the data set D is normalized so that each element in the data set D has the same scalar, wherein the data set D = [a, b, c, d]; Among them, a is the normalized historical payment record sequence of the passenger, b is the normalized credit score, c is the normalized blacklist record ratio, and d is the normalized fraud record ratio.

3. The bad debt risk prediction method according to claim 2, characterized in that: In S2, the execution process of the DS evidence theory algorithm includes: S200, set the risky proposition H1 and the risk-free proposition H2; S201, based on the data set D, decompose it into two evidence sets A and B corresponding to the risky proposition H1 and the risk-free proposition H2; S202, assign basic probability assignments BPA to the risky proposition H1 and the risk-free proposition H2, denoted as mH1(A) and mH2(B) respectively; S203, combining the retrieved passenger information and combining the evidence according to the merging principle of the DS evidence theory algorithm; S204, according to the combined probability assignment m12(E), the risk factor E can be quantified as the ratio of the probability of the risky proposition H1 to the probability of the risk-free proposition H2.

4. The bad debt risk prediction method according to claim 3, characterized in that: In S201, the decomposition method includes: S2010, for each selected feature of the data set D, setting one or more thresholds to distinguish between risky and non-risky situations; S2011, according to the set threshold, divides each sample in the data set D into evidence supporting the risky proposition H1 or the risk-free proposition H2 by creating a binary vector; 1 indicates support for the proposition and 0 indicates non-support; S2012, all samples supporting the risky proposition H1 form evidence set A, and all samples supporting the risk-free proposition H2 form evidence set B.

5. The bad debt risk prediction method according to claim 3, characterized in that: In S202, a bell-shaped function is used to assign basic probability assignments BPA to the risky proposition H1 and the risk-free proposition H2: Among them, BPA (x i ) indicates that in feature x i BPA value; x is a feature value in the data set D. μ is the center point of the bell-shaped function, indicating the most likely value or the most trusted point; α is the width parameter of the bell-shaped function, which controls the steepness of the function.

6. The bad debt risk prediction method according to claim 3, characterized in that: In S203, the combined probability assignment is expressed as m12(E), where E is the set of all possible propositions on the recognition frame: Among them, K is the conflict coefficient, which indicates the degree of inconsistency between evidences; m1(A i ) and m2(B j ) are the basic probability assignments BPA of the corresponding propositions in the evidence sets A and B respectively; A i and B j are subsets of evidence sets A and B respectively.

7. The bad debt risk prediction method according to claim 1, characterized in that: In said S4, it includes: S400, traversing the payment records of the passengers and marking all payment timeout situations; S401, collect statistics on passengers' payment frequency, payment amount distribution and common payment channels; evaluate whether passengers' payment habits fluctuate or are inconsistent; S402, compare the passenger's payment record with the fraud records in the external black market data source; check whether the passenger's payment account, payment channel and payment IP address appear in the fraud database; once the passenger is found to be involved in fraud, identify the risk.

8. The bad debt risk prediction method according to claim 1, 2, 3 or 7, characterized in that: Also includes: S5, generating a risk assessment report based on the results of risk assessment and detailed payment information retrieval; Including passengers' bad debt warning information and existing risk points.

9. A system for implementing the bad debt risk prediction method according to any one of claims 1 to 8, characterized in that: The system comprises: Receive personal information submitted by passengers when placing orders, and access the data receiving and pre-processing module of the risk control database of the designated driver service system; The DS evidence theory algorithm is introduced to calculate the risk factor E risk assessment and calculation module; A risk assessment and prevention module that compares the risk factor E with the preset threshold T to determine the bad debt risk of the passenger's order; A payment status retrieval module for retrieving a passenger's past payment status.

10. The system according to claim 9, characterized in that: It also includes a risk assessment result output module that generates a risk assessment report based on the results of risk assessment and detailed payment situation retrieval.

Images