Method and system for monitoring and identifying abnormal data of factoring service

Abstract

The invention provides a factoring business abnormal data monitoring identification method and system, and relates to the technical field of factoring business management.The method comprises the steps that multiple factoring business monitoring factors are configured, a factoring business monitoring tree is constructed, a target user is monitored in real time, a factoring business monitoring data tree is obtained, and a factoring business abnormal detection component is constructed. And constructing a factoring service exception combing function, obtaining a factoring service exception monitoring report, and performing graded early warning. Through the factoring service risk management method and device, the technical problem of poor factoring service risk management quality caused by single factoring service monitoring factor and low anomaly detection precision in the prior art can be solved, the factoring service monitoring data tree is obtained through real-time monitoring of a user by constructing multiple factoring service monitoring factors, and the factoring service risk management quality is improved. And performing anomaly identification and early warning according to the anomaly detection component and the anomaly sorting function, thereby achieving the technical effect of improving the factoring business risk management quality.

Description

Technical Field

[0001] The present application relates to the technical field of factoring business management, and in particular to a method and system for monitoring and identifying abnormal data in factoring business. Background Art

[0002] With the rapid development of the digital economy, factoring business is undergoing a transformation from traditional financing to "financing + service". The application of cutting-edge technologies such as big data, artificial intelligence, and blockchain has promoted the scenario-based, online, and data-based development of factoring business. With the continuous growth of factoring business volume, how to efficiently process and analyze large amounts of data has become a challenge. At present, the existing indicators of factoring business management are not comprehensive enough, and the monitoring system often relies on regularly updated data.

[0003] In summary, the existing technology has the technical problem that the quality of risk management of factoring business is poor due to the single monitoring factor and low accuracy of anomaly detection of factoring business. Summary of the invention

[0004] The purpose of this application is to provide a method and system for monitoring and identifying abnormal data in factoring business, so as to solve the technical problem in the prior art that the quality of risk management of factoring business is poor due to the single monitoring factor and low accuracy of abnormal detection in factoring business.

[0005] In view of the above problems, the present application provides a method and system for monitoring and identifying abnormal data in factoring business.

[0006] In the first aspect, the present application provides a method for monitoring and identifying abnormal data of factoring business, and the method is implemented by a system for monitoring and identifying abnormal data of factoring business, wherein the method includes: configuring multiple factoring business monitoring factors, wherein the multiple factoring business monitoring factors include asset quality factors, solvency factors, market factors and business activity factors; constructing a factoring business monitoring tree according to the multiple factoring business monitoring factors; performing real-time monitoring of target users according to the factoring business monitoring tree to obtain a factoring business monitoring data tree; building a factoring business anomaly detection component according to the factoring business monitoring tree; constructing a factoring business anomaly combing function; performing anomaly identification on the factoring business monitoring data tree according to the factoring business anomaly detection component and the factoring business anomaly combing function to obtain a factoring business anomaly monitoring report; and performing graded warnings in combination with the factoring business anomaly monitoring report according to a factoring business anomaly warning device.

[0007] In the second aspect, the present application also provides an abnormal data monitoring and identification system for factoring business, which is used to execute an abnormal data monitoring and identification method for factoring business as described in the first aspect, wherein the system includes: a monitoring factor configuration module, the monitoring factor configuration module is used to configure multiple factoring business monitoring factors, wherein the multiple factoring business monitoring factors include asset quality factors, solvency factors, market factors and business activity factors; a monitoring tree construction module, the monitoring tree construction module is used to construct a factoring business monitoring tree based on the multiple factoring business monitoring factors; a real-time monitoring module, the real-time monitoring module is used to monitor the target users based on the factoring business monitoring tree. Conduct real-time monitoring to obtain a factoring business monitoring data tree; an anomaly detection module, the anomaly detection module is used to build a factoring business anomaly detection component according to the factoring business monitoring tree; an anomaly combing function construction module, the anomaly combing function construction module is used to construct a factoring business anomaly combing function; an anomaly identification module, the anomaly identification module is used to perform anomaly identification on the factoring business monitoring data tree according to the factoring business anomaly detection component and the factoring business anomaly combing function, and obtain a factoring business anomaly monitoring report; an anomaly warning module, the anomaly warning module is used to perform graded warning according to the factoring business anomaly warning device and the factoring business anomaly monitoring report.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages: By configuring multiple factoring business monitoring factors, wherein the multiple factoring business monitoring factors include asset quality factors, solvency factors, market factors and business activity factors; constructing a factoring business monitoring tree according to the multiple factoring business monitoring factors; conducting real-time monitoring of target users according to the factoring business monitoring tree to obtain a factoring business monitoring data tree; building a factoring business anomaly detection component according to the factoring business monitoring tree; constructing a factoring business anomaly combing function; identifying anomalies of the factoring business monitoring data tree according to the factoring business anomaly detection component and the factoring business anomaly combing function to obtain a factoring business anomaly monitoring report; and conducting graded warnings in combination with the factoring business anomaly monitoring report according to the factoring business anomaly warning device. In other words, by constructing multiple factoring business monitoring factors, monitoring users in real time to obtain a factoring business monitoring data tree, and performing anomaly identification and warnings according to the anomaly detection component and the anomaly combing function, the technical effect of improving the quality of factoring business risk management is achieved.

[0009] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented according to the contents of the specification, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are specifically cited below. It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easy to understand through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions in the present application or 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 in the following description are only exemplary, and for ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0011] Figure 1 A flowchart of a method for monitoring and identifying abnormal data in factoring business is provided for this application; Figure 2 This is a structural diagram of an abnormal data monitoring and identification system for factoring business in the present application.

[0012] Explanation of the accompanying drawings: monitoring factor configuration module 11, monitoring tree construction module 12, real-time monitoring module 13, anomaly detection module 14, anomaly combing function construction module 15, anomaly identification module 16, anomaly warning module 17. DETAILED DESCRIPTION

[0013] This application solves the technical problem in the prior art that the quality of risk management of factoring business is poor due to the single monitoring factor and low anomaly detection accuracy of factoring business by providing a method and system for monitoring and identifying abnormal data of factoring business. By constructing multiple factoring business monitoring factors, real-time monitoring of users is performed to obtain a factoring business monitoring data tree, and anomaly identification and early warning are performed based on anomaly detection components and anomaly combing functions, the technical effect of improving the quality of factoring business risk management is achieved.

[0014] Below, the technical solutions in the present application will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments of the present application. It should be understood that the present application is not limited to the example embodiments described herein. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present application. It should also be noted that, for the convenience of description, only the parts related to the present application are shown in the accompanying drawings, rather than all of them. Example

[0015] Please see attached Figure 1 The present application provides a method for monitoring and identifying abnormal data of factoring business, wherein the method is applied to a system for monitoring and identifying abnormal data of factoring business, and the method specifically comprises the following steps: Step 1: Configure multiple factoring business monitoring factors, wherein the multiple factoring business monitoring factors include asset quality factors, solvency factors, market factors and business activity factors.

[0016] Specifically, a comprehensive set of monitoring factors is configured by comprehensively considering the characteristics and risk sources of factoring business, including asset quality factors, solvency factors, market factors and business activity factors. Asset quality factors mainly focus on the quality of the claims and the credit status of the debtors in the factoring business, including the credit rating of the debtors and the guarantee of the claims; solvency factors evaluate the financial status and repayment ability of the debtors, including their cash flow status and financial status; market factors involve external factors such as the macroeconomic environment and industry development trends, such as GDP growth rate and industry growth rate; business activity factors focus on the operation of the factoring business itself, such as transaction frequency and transaction scale. Configuring multiple factoring business monitoring factors and building a multi-dimensional risk monitoring framework can capture the risks of factoring business operations from different angles and levels. For example, by monitoring asset quality factors, we can timely discover the deterioration of the debtor's credit status and take immediate measures to reduce risks; monitoring solvency factors can predict the debtor's solvency and reduce risks in advance; monitoring market factors can grasp the external risks faced by factoring business from a macro perspective and adjust business strategies in a timely manner; monitoring business activity factors can observe the operating status of factoring business from a micro perspective and take measures to detect abnormalities in a timely manner. By configuring multiple factoring business monitoring factors, a comprehensive and dynamic risk monitoring system can be established to improve the accuracy of risk identification.

[0017] Step 2: Construct a factoring business monitoring tree based on the multiple factoring business monitoring factors.

[0018] Specifically, these monitoring factors are organized according to a certain logic and structure to form a hierarchical and well-organized framework. Consider the characteristics of individual factors and determine the weights and relationships of each monitoring factor. For example, the asset quality factor is given a higher weight because it is directly related to the security of the factoring assets, while the market factor has a lower weight because it has a smaller impact on the overall risk. After determining the weights and relationships, a monitoring tree is constructed. The root node is usually an overall risk indicator, such as the overall risk level of the factoring business. The value of the root node depends on the value of its child nodes, that is, the comprehensive performance of other monitoring factors. The first-level leaf nodes are different monitoring factors, such as asset quality factors, solvency factors, etc. The second-level leaf nodes are important risk factors or indicators related to each monitoring factor, such as the debtor's credit rating, the company's financial status, etc. By constructing a monitoring tree, we can have a more comprehensive understanding of the risk status of the factoring business, discover potential risk factors in a timely manner, and take corresponding measures to control risks. When the value of a first-level leaf node exceeds a certain threshold, we can further analyze its subordinate second-level nodes to find out the specific source of risk and formulate corresponding strategies. For example, when the value of the asset quality factor exceeds a certain threshold, further analyze its subordinate indicators such as overdue rate and bad debt rate to find out the specific source of risk and formulate corresponding risk management strategies. Constructing a rough tree for factoring business monitoring is an effective risk management tool that can better understand the risk status of factoring business, timely discover potential risk factors, and take corresponding measures for risk control.

[0019] Step 3: According to the factoring business monitoring tree, real-time monitoring is performed on the target user to obtain a factoring business monitoring data tree.

[0020] Specifically, real-time monitoring of the target user's business activities, collection of relevant data and information, ensure that the factoring company can promptly identify potential risk factors and take preventive or corrective measures. Determining the method and frequency of data collection usually relies on advanced data acquisition and processing technology, which can quickly and accurately acquire and process a large amount of business data. For example, by accessing the user's financial system, transaction system, etc., the user's transaction data, financial data and other information can be acquired in real time. The data is processed and analyzed to generate a factoring business monitoring data tree, which includes the user's basic information, transaction information, financial information, etc. For example, if one of the indicators in the monitoring tree is the overdue rate of accounts receivable, the factoring company needs to collect the target user's accounts receivable data and calculate the proportion of overdue accounts, that is, the value of the corresponding node in the monitoring data tree. If the factoring company finds through real-time monitoring that the overdue rate of accounts receivable of a certain user has suddenly increased, indicating that its financial status is deteriorating, the factoring company needs to take immediate measures, such as adjusting the credit limit for the user, requiring additional guarantees or even suspending further financing. Real-time monitoring of target users is carried out to obtain a factoring business monitoring data tree, enabling factoring companies to make decisions based on real-time data, thereby improving the timeliness and accuracy of risk management.

[0021] Step 4: Build a factoring business anomaly detection component based on the factoring business monitoring tree.

[0022] Specifically, the key indicators and thresholds for anomaly detection are determined according to the factoring business monitoring tree, and which situations are considered abnormal. By setting the threshold, that is, for each monitoring indicator, a normal range is determined, and data outside this range will be considered abnormal. The setting of key indicators and thresholds is based on a deep understanding of the factoring business, and can be determined based on historical data, industry standards, and expert opinions to reflect the normal and abnormal states in the operation of the business. For example, transaction frequency, transaction amount, debtor credit rating, etc. may be selected as key indicators, and specific thresholds are used to distinguish between normal and abnormal situations. Select appropriate algorithms and models to build anomaly detection components, which can be statistical models, machine learning models, or deep learning models, etc., to analyze monitoring data and identify abnormal patterns that are significantly different from normal patterns. For example, clustering algorithms are used to identify abnormal clusters in data, or time series analysis is used to detect abnormal trends in data. Building factoring business anomaly detection components can realize automatic detection of abnormal situations in factoring business, reduce the need for manual monitoring and data inspection, reduce the possibility of human errors, and work uninterruptedly. For example, suppose a factoring company sets a threshold for the overdue rate of accounts receivable. When the overdue rate of any corporate customer exceeds this threshold, the anomaly detection component will detect it immediately and automatically send an alert to the risk management team. After receiving the alert, the risk management team will quickly investigate the cause and decide whether to take measures, such as adjusting the credit limit, requesting additional guarantees, or suspending financing. For another example, if the monitoring indicators of market factors suddenly change significantly, such as the industry growth rate or the degree of market competition, the anomaly detection component can also capture these changes in time, reminding the factoring company that it needs to re-evaluate the credit risk of the entire industry. Through automation and intelligent means, the efficiency of risk identification and response is improved.

[0023] Step 5: Construct a function to sort out factoring business exceptions.

[0024] Specifically, by combining multiple key risk factors, a comprehensive abnormal risk coefficient is calculated to comprehensively evaluate the abnormal risks in factoring business. The risk prediction coefficients are determined, including the asset quality abnormal risk prediction coefficient corresponding to the asset quality factor, the solvency abnormal risk prediction coefficient corresponding to the solvency factor, the market abnormal risk prediction coefficient corresponding to the market factor, and the business activity abnormal risk prediction coefficient corresponding to the business activity factor. The abnormal risk coefficient of factoring business is calculated to reflect the potential risk level of different risk areas, which can more accurately evaluate and compare the importance of different risk areas and their impact on the overall risk.

[0025] Step 6: Anomaly identification is performed on the factoring business monitoring data tree according to the factoring business anomaly detection component and the factoring business anomaly combing function to obtain a factoring business anomaly monitoring report.

[0026] Specifically, the factoring business anomaly detection component is used to perform real-time analysis on the monitoring data tree to identify and analyze abnormal patterns in the data. The factoring business anomaly combing function is used to calculate the detected anomalies to obtain a comprehensive anomaly risk coefficient. Based on the output of the anomaly detection component and the calculation results of the anomaly combing function, anomalies in the factoring business are identified, including the debtor's credit rating decline, abnormal transaction frequency, etc. The identified anomalies are organized into a factoring business anomaly monitoring report, including information such as the type of anomaly, the time of occurrence, the amount involved, etc., and the anomalies found during the monitoring process are recorded in detail. By generating a monitoring report, important data support is provided for the risk management of insurance companies, and anomalies in factoring business can be effectively discovered and handled.

[0027] Step 7: According to the factoring business abnormality warning device, a graded warning is performed in combination with the factoring business abnormality monitoring report.

[0028] Specifically, according to the characteristics of factoring business and risk management objectives, different warning levels and thresholds are set, and factoring business abnormal warning devices are configured. Analyze the abnormal monitoring report of factoring business and identify abnormal situations of different levels. Determine the warning level corresponding to each abnormal situation based on the configuration of the warning device and the analysis results of the abnormal monitoring report. The warning level is divided according to the severity of the abnormality and the scope of impact, such as low-level warning, medium-level warning and high-level warning. According to the warning level, formulate corresponding warning measures, including adjusting credit strategies, warning notifications, etc. According to the warning measures, implement corresponding warning operations, issue warning notifications to relevant personnel, initiate risk disposal procedures, etc. According to the abnormal warning device of factoring business, combined with the abnormal monitoring report of factoring business, graded warning can more effectively deal with abnormal situations of different levels and ensure the stable operation of factoring business.

[0029] Furthermore, step 2 of this application includes: Monitoring indicator mining is performed according to the multivariate factoring business monitoring factors to obtain a multivariate factoring business monitoring indicator set; a factoring business monitoring rough tree is established with the multivariate factoring business monitoring factors as the first-level leaf nodes and the multivariate factoring business monitoring indicator set as the second-level leaf nodes; a second-level leaf node similarity analysis is performed on the factoring business monitoring rough tree to obtain a second-level leaf node similarity analysis result; based on a predetermined leaf node similarity and in combination with the second-level leaf node similarity analysis result, the second-level leaf node optimization is performed on the factoring business monitoring rough tree to obtain the factoring business monitoring tree.

[0030] Specifically, various data related to factoring business are collected, including corporate financial statements, market research reports, industry data, etc. Data mining technology and statistical analysis methods are used to conduct in-depth analysis of these data. Monitoring indicator mining includes association rule mining, cluster analysis, factor analysis, etc. Association rule mining can find the correlation between different indicators, for example, it is found that there is a correlation between the overdue rate of accounts receivable and the financial status of the enterprise; cluster analysis groups similar enterprises or indicators to better understand the relationship between them; factor analysis can identify the main influencing factors, thereby establishing a more streamlined indicator system. The multi-factoring business monitoring factors are taken as the first-level leaf nodes, including asset quality factors, solvency factors, market factors and business activity factors, representing different aspects that affect the risk of factoring business. The second-level leaf nodes include specific monitoring indicators, such as the debtor's credit rating, transaction amount, transaction frequency, etc., which are based on the further refinement and concretization of the first-level factors.

[0031] Perform similarity analysis on the secondary leaf nodes, compare and evaluate the various indicators in the monitoring indicator set to determine the degree of similarity between them. It usually relies on advanced algorithms and models, such as calculating the correlation, covariance or other statistical methods between indicators, to identify which indicators in the monitoring indicator set have high similarity, that is, which indicators have overlapping or redundant information in reflecting risks. When two indicators show a high positive correlation in historical data, that is, when one indicator rises, the other indicator also rises. It is identified that the two indicators are similar in the direction of monitoring risk, thereby reducing repeated monitoring. After obtaining the similarity analysis results of the secondary leaf nodes, similar indicators are merged into a comprehensive indicator, or the weights of certain indicators are adjusted to improve the accuracy and efficiency of monitoring. For example, assuming that the similarity between the overdue rate and the bad debt rate of accounts receivable exceeds the predetermined threshold, that is, the two indicators have a high degree of similarity in risk monitoring, you can choose to retain one of the indicators, such as the overdue rate of accounts receivable, and remove or merge the bad debt rate into the overdue rate of accounts receivable. In this way, you can focus more on analyzing those truly independent indicators that can provide additional risk insights. By mining monitoring indicators, establishing a rough monitoring tree, and conducting similarity analysis and optimization of secondary leaf nodes, a more accurate and efficient factoring business monitoring tree is constructed, thereby enhancing the accuracy of risk monitoring.

[0032] Furthermore, the present application also includes the following steps: Obtain a factoring business monitoring event record set; perform feature identification based on the factoring business monitoring event record set to obtain a factoring business monitoring sample indicator set; perform association selection on the factoring business monitoring sample indicator set based on the asset quality factor to obtain an asset quality monitoring indicator set; perform decoupling optimization based on the asset quality monitoring indicator set to obtain an optimized asset quality monitoring indicator set, and add the optimized asset quality monitoring indicator set to the multi-factoring business monitoring indicator set; based on the factoring business monitoring event record set, continue to perform monitoring indicator mining on the multi-factoring business monitoring factors to obtain the multi-factoring business monitoring indicator set.

[0033] Specifically, monitoring indicators are identified and optimized through feature identification, association selection, decoupling optimization, and monitoring indicator mining. First, detailed information on monitoring events related to factoring business is collected from multiple channels to obtain a factoring business monitoring event record set, including transaction data, debtor information, market data, etc. These data are collected from the internal systems of the factoring company, such as the financial system, transaction system, etc., or from external data sources, such as credit rating agencies, market research institutions, etc. For example, a factoring business monitoring event record set may contain information such as the debtor's credit rating, transaction amount, transaction frequency, and debtor's industry background. Next, feature identification is performed based on the factoring business monitoring event record set, and the data in the event record set is analyzed to extract features that are meaningful for risk assessment from a large amount of data. These features can reflect the key factors of factoring business risk, such as overdue rate, bad debt rate, corporate financial ratio, etc. Through feature identification, complex data sets can be converted into indicator sets that can be used for risk analysis. For example, by analyzing the debtor's credit score and transaction amount, an indicator set that reflects the debtor's solvency and credit status can be constructed.

[0034] After obtaining the sample indicator set, the indicators related to asset quality are screened out from the sample indicator set for factoring business monitoring to obtain the asset quality monitoring indicator set. Asset quality factors include the credit status of the debtor, the security of the claims, etc., which are directly related to the security of the factoring assets. Statistical methods or machine learning algorithms are used to evaluate the degree of correlation between each indicator in the sample indicator set for factoring business monitoring and the asset quality factors. According to the evaluation results, the indicators with a high degree of correlation with the asset quality factors are screened out. The screened indicators are optimized to detect redundant or highly correlated indicators, which come from different but substantially similar monitoring sources or have limited contribution in risk assessment. These indicators are processed, including merging, deleting, and adjusting weights, and this optimized indicator set is then added to the multi-factoring business monitoring indicator set. For example, in the optimized indicator set, if the two indicators of the debtor's credit score and financial leverage ratio are included, there may be a certain degree of duplication between the two indicators in reflecting the debtor's debt repayment ability. Through decoupling optimization, it can be decided to retain only one of the indicators, thereby reducing the monitoring burden and improving the efficiency of monitoring. Finally, we conduct a deeper analysis of the factoring business monitoring event record set to identify more characteristics and indicators related to factoring business risks, and extract meaningful characteristics for risk assessment from the data obtained from the in-depth analysis, including the debtor's credit status, solvency, market environment, business activities, etc. Through continuous monitoring indicator mining, we continuously update and improve the monitoring indicator set to adapt to market changes and business needs.

[0035] Feature identification refers to the process of identifying indicators that are meaningful for factoring business risk monitoring from a large amount of data, reflecting the key factors of factoring business risk, such as overdue rate, bad debt rate, corporate financial ratio, etc. Association selection is to screen the sample indicator set according to a specific monitoring factor, such as asset quality factor, and select indicators that are highly correlated with the factor in order to more accurately monitor and evaluate the risks associated with the factor. For example, under the asset quality factor, it is highly correlated with indicators such as overdue rate and bad debt rate. Through association selection, those indicators that are not closely related to specific monitoring factors are removed, thereby improving the pertinence of risk monitoring. Decoupling optimization is to further analyze and process the selected indicators on the basis of association selection, remove the redundancy and dependence between indicators, so that each indicator can independently reflect the risk. Monitoring indicator mining is to continue to conduct data mining on multiple factoring business monitoring factors after obtaining a preliminary monitoring indicator set, so as to discover more monitoring indicators, ensure the comprehensiveness of the monitoring indicator set, and cover all aspects of factoring business. Monitoring indicator mining can adopt various data mining techniques and statistical analysis methods, such as association rule mining, cluster analysis, factor analysis, etc. By obtaining a diversified set of factoring business monitoring indicators, the efficiency of risk management can be improved while removing redundant indicators.

[0036] Furthermore, the present application also includes the following steps: According to the asset quality monitoring indicator set, pairwise coupling identification is performed to obtain multiple monitoring indicator coupling coefficients; based on the multiple monitoring indicator coupling coefficients and in combination with the monitoring indicator coupling threshold, the asset quality monitoring indicator set is selected to obtain multiple identification monitoring indicator groups that are greater than / equal to the monitoring indicator coupling threshold; indicators are fused for the multiple identification monitoring indicator groups respectively to obtain multiple decoupled monitoring indicators; the asset quality monitoring indicator set is corrected according to the multiple decoupled monitoring indicators to obtain the optimized asset quality monitoring indicator set.

[0037] Specifically, the indicators in the asset quality monitoring indicator set are paired, and coupling analysis is performed on each pair of indicators in the monitoring indicator set, including the debtor's credit score and financial leverage ratio, transaction amount and transaction frequency, etc. The coupling degree of each identified pair of coupling indicators is calculated through statistical methods or machine learning algorithms, such as Pearson correlation coefficient, multivariate regression analysis, etc., to measure the correlation and dependence between the two indicators. Based on the calculation results, the coupling coefficient between each pair of indicators is determined. The coupling coefficient reflects the degree of association between the two indicators, usually expressed in numerical form. The higher the value, the stronger the correlation between the two indicators. According to business needs and risk management objectives, based on historical data, industry standards or expert opinions, a monitoring indicator coupling threshold is set to determine whether the coupling degree between the two indicators is high enough. According to the coupling coefficient and the set threshold, a pair of indicators with a coupling coefficient greater than / equal to the threshold is screened out, and all pairs of indicators that meet the conditions are merged into a comprehensive indicator, that is, multiple identification monitoring indicator groups. Each group consists of a group of highly correlated indicators that provide similar information in risk monitoring. Indicator fusion for each identification monitoring indicator group means that the indicators within the group need to be merged or simplified to reduce redundancy and improve efficiency. Indicator fusion can be achieved in many ways, such as principal component analysis, weighted average or other synthesis methods. After fusion, multiple decoupled monitoring indicators are obtained, which are the results of optimization. While maintaining the original risk information, data redundancy and complexity are reduced. Finally, according to these decoupled monitoring indicators, corrections are made, including adjusting the weights of indicators, merging similar indicators or eliminating irrelevant indicators, to obtain an optimized asset quality monitoring indicator set.

[0038] Pairwise coupling identification refers to analyzing each indicator in the factoring business monitoring indicator set one by one to determine the correlation or dependency with other indicators. Selection is the process of screening the indicators obtained by pairwise coupling identification based on the monitoring indicator coupling threshold to determine which indicators are necessary in risk monitoring and which may be redundant. Indicator fusion refers to the process of merging or simplifying indicators within the identification monitoring indicator group, which can reduce redundant information while retaining key risk information. For example, a company collected financial statements, market research reports and industry data from corporate customers, analyzed and identified the overdue rate of accounts receivable, the current ratio of the enterprise, the growth rate of the industry and the market share of the enterprise. The calculation showed that the coupling coefficient between the overdue rate of accounts receivable and the current ratio of the enterprise was high. The company set the coupling threshold to 0.8 and selected indicators greater than or equal to 0.8. The coupling coefficient between the overdue rate of accounts receivable and the bad debt rate exceeded 0.8, providing similar information in risk monitoring, and the company chose to retain one of them. According to the coupling threshold of monitoring indicators, highly related indicators are grouped to form multiple identification monitoring indicator groups. One group includes the overdue rate and bad debt rate of accounts receivable, and the other group includes the current ratio and debt-to-asset ratio of the enterprise. The overdue rate of accounts receivable and the current ratio of the enterprise are fused by weighted average method to obtain a comprehensive indicator reflecting the debt repayment ability and asset quality of the enterprise. According to the fused indicators, the company found that the current ratio and debt-to-asset ratio of the enterprise were redundant in reflecting risks, and chose to retain one of them to simplify the monitoring system. By reducing redundant indicators and building a more accurate and efficient monitoring system, the efficiency of risk management is improved.

[0039] Further, step 4 of this application includes: According to the factoring business monitoring tree, an asset quality anomaly detection channel is constructed; according to the factoring business monitoring tree, a solvency anomaly detection channel is constructed; according to the factoring business monitoring tree, a market anomaly detection channel is constructed; according to the factoring business monitoring tree, a business activity anomaly detection channel is constructed; the asset quality anomaly detection channel, the solvency anomaly detection channel, the market anomaly detection channel and the business activity anomaly detection channel are connected to generate the factoring business anomaly detection component.

[0040] Specifically, the factoring business monitoring tree is analyzed to identify the monitoring nodes related to asset quality. First, according to the factoring business monitoring tree, the monitoring leaf nodes related to asset quality are extracted, and the factoring business monitoring records of these leaf nodes are retrieved to obtain the asset quality monitoring record set, the asset quality anomaly identification record set and the asset quality anomaly risk coefficient record set. The asset quality anomaly identification model is generated by using the asset quality monitoring record set and the asset quality anomaly identification record set. The asset quality anomaly identification record set and the asset quality anomaly risk coefficient record set are used to generate the asset quality anomaly risk prediction model. The asset quality anomaly identification model and the asset quality anomaly risk prediction model are connected to form a complete asset quality anomaly detection channel, which can automatically receive and process monitoring data and compare the data with the set threshold. Once the data exceeds the threshold, the system should automatically trigger the alarm mechanism to notify the risk management personnel to take further action. Similarly, in the same way, connect the solvency anomaly identification record set and the solvency anomaly risk coefficient record set to generate a solvency anomaly detection channel; connect the market anomaly identification record set and the market anomaly risk coefficient record set to generate a market anomaly detection channel; connect the business activity anomaly identification record set and the business activity anomaly risk coefficient record set to generate a business activity anomaly detection channel. Integrate the above four anomaly detection channels, design a factoring business anomaly detection component based on business needs and risk management goals, comprehensively cover the risk monitoring of factoring business, and can perform layered processing according to the severity of the risk. By generating a factoring business anomaly detection component, real-time monitoring and timely response to risks are ensured.

[0041] Furthermore, the present application also includes the following steps: Extract asset quality monitoring leaf nodes according to the factoring business monitoring tree; retrieve factoring business monitoring records according to the asset quality monitoring leaf nodes to obtain asset quality monitoring record sets, asset quality anomaly identification record sets and asset quality anomaly risk coefficient record sets; train a predetermined learning channel according to the asset quality monitoring record sets and the asset quality anomaly identification record sets, and obtain a first mean square loss error every time the training is performed for a predetermined number of times; if the first mean square loss error is less than a predetermined mean square loss error threshold, generate an asset quality anomaly identification model; train the predetermined learning channel according to the asset quality anomaly identification record set and the asset quality anomaly risk coefficient record set, and obtain a second mean square loss error every time the training is performed for the predetermined number of times; if the second mean square loss error is less than the predetermined mean square loss error threshold, generate an asset quality anomaly risk prediction model; connect the asset quality anomaly identification model and the asset quality anomaly risk prediction model to generate the asset quality anomaly detection channel.

[0042] Specifically, the monitoring leaf nodes related to asset quality represent key indicators of asset quality monitoring, such as the overdue rate of accounts receivable, bad debt rate, and financial status of the enterprise. Using the extracted asset quality monitoring leaf nodes, the relevant asset quality monitoring record set is retrieved and sorted from the factoring business monitoring records, which contains all monitoring data related to asset quality; the asset quality monitoring record set is further analyzed, and abnormal patterns are monitored through statistical methods or machine learning algorithms, and the identified abnormalities are recorded in the asset quality abnormality identification record set; based on the asset quality abnormality identification record set, the asset quality abnormality risk coefficient is calculated to reflect the risk level of asset quality abnormalities, and these coefficients are recorded in the asset quality abnormality risk coefficient record set. Select a suitable machine learning model to build a predetermined learning channel, and you can choose neural networks, decision trees, support vector machines, random forests, etc.

[0043] Use the asset quality monitoring record set and the asset quality anomaly identification record set to train the selected model, and continuously adjust its internal parameters to minimize the prediction error. Select a suitable loss function to measure the prediction error of the model. For example, use the mean square error (MSE) as the loss function to measure the difference between the model prediction value and the true value. Set the number of model training times, that is, the number of times the training data set is repeated during the training process. After each predetermined number of training, obtain the first mean square loss error to evaluate the performance of the model. According to business needs and model performance, set a predetermined mean square loss error threshold. This threshold is set based on historical data or expert experience and is used to judge the critical value of the model training effect. During the model training process, the first mean square loss error is calculated regularly and compared with the predetermined mean square loss error threshold. If the first mean square loss error is less than the predetermined mean square loss error threshold, it is considered that the model has achieved the predetermined training effect and can be used to identify asset quality anomalies and generate an asset quality anomaly identification model.

[0044] Then, the predetermined learning channel is trained using the asset quality anomaly identification record set and the asset quality anomaly risk coefficient record set. During each training process, the factoring company will obtain the second mean square loss error. If the second mean square loss error is less than the predetermined mean square loss error threshold, it is considered that the model has been trained well enough to be used for the prediction of asset quality anomaly risk and generate an asset quality anomaly risk prediction model. Connecting the asset quality anomaly identification model and the asset quality anomaly risk prediction model involves combining the outputs of the two models or using the output of one model as the input of another model. According to business needs and risk management objectives, an asset quality anomaly detection channel is designed to cover all aspects of asset quality monitoring and be able to perform layered processing according to the severity of the risk. The constructed asset quality anomaly detection channel is verified to ensure that it can accurately identify asset quality anomalies and can operate effectively in actual business. According to the verification results and actual business needs, the asset quality anomaly detection channel is optimized, including adjusting model parameters, adding new monitoring indicators, or adjusting the structure of the channel. Connecting the asset quality anomaly identification model and the asset quality anomaly risk prediction model to generate an asset quality anomaly detection channel will help build a comprehensive and efficient asset quality anomaly detection system, thereby ensuring the stable operation of factoring business.

[0045] Further, step five of this application includes: The factoring business exception sorting function is: in, Characterize the abnormal risk coefficient of factoring business, Characterizes the asset quality abnormal risk prediction coefficient, Characterizes the solvency abnormal risk prediction coefficient, Characterizes the market abnormal risk prediction coefficient, Characterizes the abnormal risk prediction coefficient of business activities.

[0046] Specifically, the factoring business abnormality combing function is a mathematical model used to quantify the risk of factoring business. It evaluates the overall risk level of factoring business by calculating the ratio of the sum of the squares of each risk prediction coefficient to the sum of the prediction coefficients. First, determine the abnormal risk prediction coefficient for each risk area based on historical data, industry standards, expert opinions or other risk assessment models. Characterizes the asset quality abnormal risk prediction coefficient, reflecting the impact of factors such as the debtor's credit status and the guarantee status of the creditor's rights on the risk; Characterizes the abnormal risk prediction coefficient of solvency, reflecting the impact of factors such as the debtor's financial status and cash flow status on the risk; Characterizes the market abnormal risk prediction coefficient, reflecting the impact of macroeconomic environment, industry development trends and other factors on risk; The prediction coefficient that characterizes the abnormal risk of business activities reflects the operation of the factoring business itself, such as the impact of factors such as transaction frequency and transaction scale on risk. Next, the square of each prediction coefficient is added together to amplify the negative impact of the risk factor and make the model more sensitive to risk. The sum of the prediction coefficients is calculated to reflect the overall contribution to the risk. Finally, the abnormal risk coefficient of the factoring business is calculated by dividing the square of each risk factor by the sum of the prediction coefficients. It is a value between 0 and 1. The closer it is to 1, the higher the risk and the more attention it needs. This function can more accurately assess and compare the importance of different risk areas and their impact on the overall risk.

[0047] In summary, the abnormal data monitoring and identification method for factoring business provided by this application has the following technical effects: By configuring multiple factoring business monitoring factors, wherein the multiple factoring business monitoring factors include asset quality factors, solvency factors, market factors and business activity factors; constructing a factoring business monitoring tree according to the multiple factoring business monitoring factors; conducting real-time monitoring of target users according to the factoring business monitoring tree to obtain a factoring business monitoring data tree; building a factoring business anomaly detection component according to the factoring business monitoring tree; constructing a factoring business anomaly combing function; identifying anomalies of the factoring business monitoring data tree according to the factoring business anomaly detection component and the factoring business anomaly combing function to obtain a factoring business anomaly monitoring report; and conducting graded warnings in combination with the factoring business anomaly monitoring report according to the factoring business anomaly warning device. In other words, by constructing multiple factoring business monitoring factors, monitoring users in real time to obtain a factoring business monitoring data tree, and performing anomaly identification and warnings according to the anomaly detection component and the anomaly combing function, the technical effect of improving the quality of factoring business risk management is achieved. Example

[0048] Based on the same inventive concept as the abnormal data monitoring and identification method for factoring business in the aforementioned embodiment, the present application also provides an abnormal data monitoring and identification system for factoring business, please refer to the attached Figure 2 , the system comprising: The monitoring factor configuration module 11 is used to configure multiple factoring business monitoring factors, wherein the multiple factoring business monitoring factors include asset quality factors, solvency factors, market factors and business activity factors.

[0049] The monitoring tree construction module 12 is used to construct a factoring business monitoring tree according to the multiple factoring business monitoring factors.

[0050] The real-time monitoring module 13 is used to perform real-time monitoring on the target user according to the factoring business monitoring tree to obtain a factoring business monitoring data tree.

[0051] The anomaly detection module 14 is used to build a factoring business anomaly detection component according to the factoring business monitoring tree.

[0052] The abnormal combing function construction module 15 is used to construct the abnormal combing function of the factoring business.

[0053] The anomaly identification module 16 is used to identify anomalies on the factoring business monitoring data tree according to the factoring business anomaly detection component and the factoring business anomaly combing function to obtain a factoring business anomaly monitoring report.

[0054] The abnormal warning module 17 is used to make graded warnings according to the abnormal warning device for factoring business in combination with the abnormal monitoring report for factoring business.

[0055] Further, the monitoring tree construction module 12 in the system is also used for: Monitoring indicator mining is performed according to the multivariate factoring business monitoring factors to obtain a multivariate factoring business monitoring indicator set; a factoring business monitoring rough tree is established with the multivariate factoring business monitoring factors as the first-level leaf nodes and the multivariate factoring business monitoring indicator set as the second-level leaf nodes; a second-level leaf node similarity analysis is performed on the factoring business monitoring rough tree to obtain a second-level leaf node similarity analysis result; based on a predetermined leaf node similarity and in combination with the second-level leaf node similarity analysis result, the second-level leaf node optimization is performed on the factoring business monitoring rough tree to obtain the factoring business monitoring tree.

[0056] Further, the monitoring tree construction module 12 in the system is also used for: Obtain a factoring business monitoring event record set; perform feature identification based on the factoring business monitoring event record set to obtain a factoring business monitoring sample indicator set; perform association selection on the factoring business monitoring sample indicator set based on the asset quality factor to obtain an asset quality monitoring indicator set; perform decoupling optimization based on the asset quality monitoring indicator set to obtain an optimized asset quality monitoring indicator set, and add the optimized asset quality monitoring indicator set to the multi-factoring business monitoring indicator set; based on the factoring business monitoring event record set, continue to perform monitoring indicator mining on the multi-factoring business monitoring factors to obtain the multi-factoring business monitoring indicator set.

[0057] Further, the monitoring tree construction module 12 in the system is also used for: According to the asset quality monitoring indicator set, pairwise coupling identification is performed to obtain multiple monitoring indicator coupling coefficients; based on the multiple monitoring indicator coupling coefficients and in combination with the monitoring indicator coupling threshold, the asset quality monitoring indicator set is selected to obtain multiple identification monitoring indicator groups that are greater than / equal to the monitoring indicator coupling threshold; indicators are fused for the multiple identification monitoring indicator groups respectively to obtain multiple decoupled monitoring indicators; the asset quality monitoring indicator set is corrected according to the multiple decoupled monitoring indicators to obtain the optimized asset quality monitoring indicator set.

[0058] Furthermore, the anomaly detection module 14 in the system is also used for: According to the factoring business monitoring tree, an asset quality anomaly detection channel is constructed; according to the factoring business monitoring tree, a solvency anomaly detection channel is constructed; according to the factoring business monitoring tree, a market anomaly detection channel is constructed; according to the factoring business monitoring tree, a business activity anomaly detection channel is constructed; the asset quality anomaly detection channel, the solvency anomaly detection channel, the market anomaly detection channel and the business activity anomaly detection channel are connected to generate the factoring business anomaly detection component.

[0059] Furthermore, the anomaly detection module 14 in the system is also used for: Extract asset quality monitoring leaf nodes according to the factoring business monitoring tree; retrieve factoring business monitoring records according to the asset quality monitoring leaf nodes to obtain asset quality monitoring record sets, asset quality anomaly identification record sets and asset quality anomaly risk coefficient record sets; train a predetermined learning channel according to the asset quality monitoring record sets and the asset quality anomaly identification record sets, and obtain a first mean square loss error every time the training is performed for a predetermined number of times; if the first mean square loss error is less than a predetermined mean square loss error threshold, generate an asset quality anomaly identification model; train the predetermined learning channel according to the asset quality anomaly identification record set and the asset quality anomaly risk coefficient record set, and obtain a second mean square loss error every time the training is performed for the predetermined number of times; if the second mean square loss error is less than the predetermined mean square loss error threshold, generate an asset quality anomaly risk prediction model; connect the asset quality anomaly identification model and the asset quality anomaly risk prediction model to generate the asset quality anomaly detection channel.

[0060] Furthermore, the abnormal combing function construction module 15 in the system is also used for: The factoring business exception sorting function is: in, Characterize the abnormal risk coefficient of factoring business, Characterizes the asset quality abnormal risk prediction coefficient, Characterizes the solvency abnormal risk prediction coefficient, Characterizes the market abnormal risk prediction coefficient, Characterizes the abnormal risk prediction coefficient of business activities.

[0061] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. Figure 1 The abnormal data monitoring and identification method and specific examples of a factoring business in Example 1 are also applicable to an abnormal data monitoring and identification system for a factoring business in this embodiment. Through the above detailed description of the abnormal data monitoring and identification method for a factoring business, those skilled in the art can clearly know the abnormal data monitoring and identification system for a factoring business in this embodiment, so for the sake of brevity of the specification, it will not be described in detail here. For the system 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.

[0062] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

[0063] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the present application and its equivalent technology, the present application is also intended to include these modifications and variations.

Claims

1. A method for monitoring and identifying abnormal data in factoring business, characterized in that: The method comprises: Configure multiple factoring business monitoring factors, wherein the multiple factoring business monitoring factors include asset quality factors, solvency factors, market factors and business activity factors; Constructing a factoring business monitoring tree according to the multiple factoring business monitoring factors; According to the factoring business monitoring tree, real-time monitoring is performed on the target user to obtain a factoring business monitoring data tree; According to the factoring business monitoring tree, build a factoring business anomaly detection component; Construct a function to sort out abnormalities in factoring business; According to the factoring business anomaly detection component and the factoring business anomaly combing function, anomaly identification is performed on the factoring business monitoring data tree to obtain a factoring business anomaly monitoring report; According to the factoring business abnormality early warning device, a graded early warning is performed in combination with the factoring business abnormality monitoring report.

2. The method for monitoring and identifying abnormal data in factoring business according to claim 1, characterized in that: According to the multiple factoring business monitoring factors, a factoring business monitoring tree is constructed, including: Performing monitoring indicator mining according to the multi-factoring business monitoring factors to obtain a multi-factoring business monitoring indicator set; Taking the multi-factoring business monitoring factor as the first-level leaf node and the multi-factoring business monitoring indicator set as the second-level leaf node, a factoring business monitoring rough tree is established; Performing a secondary leaf node similarity analysis on the factoring business monitoring rough tree to obtain a secondary leaf node similarity analysis result; Based on the predetermined leaf node similarity, the secondary leaf node optimization is performed on the factoring business monitoring rough tree in combination with the secondary leaf node similarity analysis result to obtain the factoring business monitoring tree.

3. The method for monitoring and identifying abnormal data in factoring business according to claim 2, characterized in that: Monitoring indicator mining is performed according to the multi-factoring business monitoring factors to obtain a multi-factoring business monitoring indicator set, including: Obtain factoring business monitoring event record set; Performing feature recognition according to the factoring business monitoring event record set to obtain a factoring business monitoring sample indicator set; According to the asset quality factors, the factoring business monitoring sample indicator set is selected in association to obtain an asset quality monitoring indicator set; Performing decoupling optimization according to the asset quality monitoring indicator set to obtain an optimized asset quality monitoring indicator set, and adding the optimized asset quality monitoring indicator set to the multi-factoring business monitoring indicator set; According to the factoring business monitoring event record set, the monitoring indicator mining of the multiple factoring business monitoring factors is continued to obtain the multiple factoring business monitoring indicator set.

4. The method for monitoring and identifying abnormal data in factoring business according to claim 3, characterized in that: Decoupling optimization is performed according to the asset quality monitoring indicator set to obtain an optimized asset quality monitoring indicator set, including: Performing pairwise coupling identification according to the asset quality monitoring indicator set to obtain a plurality of monitoring indicator coupling coefficients; Based on the multiple monitoring indicator coupling coefficients, the asset quality monitoring indicator set is selected in combination with the monitoring indicator coupling threshold to obtain multiple identification monitoring indicator groups that are greater than / equal to the monitoring indicator coupling threshold; Performing indicator fusion on the multiple identification monitoring indicator groups respectively to obtain multiple decoupled monitoring indicators; The asset quality monitoring indicator set is corrected according to the multiple decoupled monitoring indicators to obtain the optimized asset quality monitoring indicator set.

5. The method for monitoring and identifying abnormal data in factoring business according to claim 1, characterized in that: According to the factoring business monitoring tree, a factoring business anomaly detection component is built, including: According to the factoring business monitoring tree, an asset quality anomaly detection channel is constructed; According to the factoring business monitoring tree, a solvency anomaly detection channel is constructed; According to the factoring business monitoring tree, a market anomaly detection channel is constructed; According to the factoring business monitoring tree, a business activity anomaly detection channel is constructed; The asset quality anomaly detection channel, the solvency anomaly detection channel, the market anomaly detection channel and the business activity anomaly detection channel are connected to generate the factoring business anomaly detection component.

6. A method for monitoring and identifying abnormal data in factoring business according to claim 5, characterized in that: According to the factoring business monitoring tree, an asset quality anomaly detection channel is constructed, including: Extracting asset quality monitoring leaf nodes according to the factoring business monitoring tree; Retrieve factoring business monitoring records according to the asset quality monitoring leaf node to obtain an asset quality monitoring record set, an asset quality anomaly identification record set, and an asset quality anomaly risk coefficient record set; Training a predetermined learning channel according to the asset quality monitoring record set and the asset quality anomaly identification record set, and obtaining a first mean square loss error after each predetermined number of trainings; If the first mean square loss error is less than a predetermined mean square loss error threshold, generating an asset quality anomaly identification model; Training the predetermined learning channel according to the asset quality anomaly identification record set and the asset quality anomaly risk coefficient record set, and obtaining a second mean square loss error each time the training is performed for the predetermined number of times; If the second mean square loss error is less than the predetermined mean square loss error threshold, generating an asset quality abnormality risk prediction model; The asset quality anomaly identification model and the asset quality anomaly risk prediction model are connected to generate the asset quality anomaly detection channel.

7. The method for monitoring and identifying abnormal data in factoring business according to claim 1, characterized in that: The factoring business exception sorting function is:

8. Among them, Characterize the abnormal risk coefficient of factoring business, Characterizes the asset quality abnormal risk prediction coefficient, Characterizes the solvency abnormal risk prediction coefficient, Characterizes the market abnormal risk prediction coefficient, Characterizes the abnormal risk prediction coefficient of business activities.

9. A system for monitoring and identifying abnormal data in factoring business, characterized in that: The system is used to implement the method for monitoring and identifying abnormal data in factoring business as described in any one of claims 1 to 7, comprising: A monitoring factor configuration module, wherein the monitoring factor configuration module is used to configure multiple factoring business monitoring factors, wherein the multiple factoring business monitoring factors include asset quality factors, solvency factors, market factors and business activity factors; A monitoring tree construction module, the monitoring tree construction module is used to construct a factoring business monitoring tree according to the multi-factoring business monitoring factors; A real-time monitoring module, which is used to perform real-time monitoring on target users according to the factoring business monitoring tree to obtain a factoring business monitoring data tree; An anomaly detection module, the anomaly detection module is used to build a factoring business anomaly detection component according to the factoring business monitoring tree; An abnormal combing function construction module, wherein the abnormal combing function construction module is used to construct an abnormal combing function for factoring business; An anomaly identification module, the anomaly identification module is used to identify anomalies on the factoring business monitoring data tree according to the factoring business anomaly detection component and the factoring business anomaly combing function, and obtain a factoring business anomaly monitoring report; The abnormal warning module is used to make graded warnings according to the abnormal warning device of the factoring business in combination with the abnormal monitoring report of the factoring business.

Images