Big data analysis method and platform for enterprise account receivable recovery risk assessment

Abstract

The invention discloses a big data analysis method and platform for enterprise account receivable recovery risk assessment, and the method comprises the steps: carrying out the analysis and feedback operation through collecting first judgment data, second judgment data, third judgment data and fourth judgment data, and judging the account receivable of an enterprise in each time period in the future and the current flow fund of the enterprise; whether the available funds of the enterprise in each time period meet the expenditure demand of the enterprise or not is predicted, so that the abnormality of the enterprise fund chain can be found in time, and enterprise management personnel can deal with the transaction money return characteristics of the same type of debt objects and the transaction money return characteristics of the debt objects themselves in real time. The method can perform early warning on the health of a debt object individual when the overall money return of an enterprise is safe, and perform early warning on the debt object according to the object category when the overall money return risk of the enterprise is relatively large.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of data analysis, and relates to a big data analysis method and platform for enterprise accounts receivable recovery risk assessment. BACKGROUND

[0002] Because of different sales modes, enterprises may not be able to receive all sales amounts in time after sales, which results in accounts receivable. In order to ensure the safety of enterprise cash flow, the quality of profit and the stability of operation, it is necessary to assess the risk of enterprise accounts receivable. The prior art with publication number CN117291603A discloses a big data comparison accounts receivable right confirmation risk assessment system, which comprises: a data acquisition unit for acquiring target enterprise data (basic information, order information, payment information, repayment information, industry information and credit rating) of an enterprise to be assessed; a data processing unit for preprocessing the target enterprise data; a feature engineering unit for determining target input features based on the preprocessed target enterprise data; and a risk assessment unit for inputting the target input features into a preset risk assessment model to obtain a risk score output by the risk assessment model, wherein the risk score is used to reveal the accounts receivable right confirmation risk of the enterprise to be assessed. The prior art considers the data characteristics in accounts receivable right confirmation by utilizing the advantages of big data, establishes a relatively comprehensive and accurate risk assessment system for accounts receivable right confirmation, and realizes risk assessment of accounts receivable right confirmation. However, whether the accounts receivable of an enterprise is healthy is not only related to other enterprise customers corresponding to the accounts receivable, but also related to the operating conditions and the income and expenditure of the enterprise itself, such as whether the current cash flow and the accounts receivable that can be recovered within a certain period meet the normal expenditure of the enterprise within the period.

[0003] The prior art does not consider the operating conditions of the enterprise itself when analyzing the health or risk of the accounts receivable of the enterprise, and only judges whether the accounts receivable is healthy according to the transactions and repayment conditions of each transaction object, which may lead to the occurrence of insufficient working capital of the enterprise, and lacks comparison of the transaction repayment characteristics of individual transaction objects corresponding to the accounts receivable of the enterprise and the overall transaction repayment characteristics of the types corresponding to the transaction objects, which is inconvenient for understanding the changes in the transaction repayment characteristics of the transaction objects and the corresponding categories. SUMMARY

[0004] In order to solve the problems existing in the prior art, the purpose of the present application is to provide a big data analysis method and platform for enterprise accounts receivable recovery risk assessment, which can predict whether the available funds of the enterprise in each period meet the expenditure demand of the enterprise by judging the accounts receivable of the enterprise in each period and the current working capital of the enterprise, so as to timely discover the abnormality of the enterprise fund chain, thereby facilitating the management personnel of the enterprise to handle it.

[0005] A big data analysis method for enterprise accounts receivable recovery risk assessment, comprising the following steps:

[0006] S1, collect enterprise accounts receivable data, enterprise planned expenditure data and enterprise current working capital data of each period in the future, analyze whether the working capital of each period in the future meets the enterprise planned expenditure data of the corresponding period, and generate first judgment data;

[0007] Among them, the enterprise accounts receivable data can include but not limited to accounts receivable unique identifier, debt object unique number, associated contract number, invoice number, accounts receivable date, agreed payment date, pre-tax receivable amount and other data, wherein the corresponding materials can be found by number, and the corresponding detailed data can be consulted.

[0008] S2, collect each debt object transaction repayment information of each period, judge whether the enterprise transaction repayment of each period is healthy, generate second judgment data; wherein the transaction repayment information is divided into transaction information and repayment information, the transaction information can include transaction serial number, transaction commodity information, transaction time, transaction responsible person, etc.; the repayment information can include repayment record unique number, actual repayment date, repayment amount, delay days, repayment bank account, etc.

[0009] S3, based on each debt object transaction repayment information of each period, generate debt object transaction repayment data, based on debt object transaction repayment data and all debt object historical transaction repayment data, analyze and process the debt object transaction repayment characteristic category corresponding to the debt object transaction repayment data, and generate object category label data;

[0010] S4, based on the debt object transaction repayment data of the current period and the corresponding debt object category label data corresponding to all other debt object transaction repayment data of the current period, judge and analyze whether the debt object transaction unit repayment data is healthy compared with the corresponding debt object category label data corresponding to all other debt object transaction repayment data, generate third judgment data;

[0011] S5, based on the current period debt object transaction repayment data and the historical period debt object transaction repayment data, judge and analyze whether the current period debt object transaction repayment data is healthy compared with the historical period debt object transaction repayment data, generate fourth judgment data;

[0012] S6, collect first judgment data, second judgment data, third judgment data and fourth judgment data for analysis feedback work.

[0013] Further, S1 includes the following steps:

[0014] S1.1, according to the time length of the account receivable time from the current time, a plurality of time periods are set, and a time period data set is generated; wherein the time length of each time period set according to actual needs, for example, the time period can be set as the time length of the account receivable time from the current time 0-3 days, 0-15 days, 0-30 days, 30-60 days, 60-90 days, 90-180 days, etc.

[0015] S1.2, collecting enterprise receivable account data, enterprise planned expenditure data and enterprise current working capital data;

[0016] S1.3, character matching search is performed on the time period data set, enterprise receivable account data and enterprise planned expenditure data, and enterprise receivable account data and enterprise planned expenditure data of each time period are searched out, and a plurality of time period enterprise receivable account data and a plurality of time period enterprise planned expenditure data are generated;

[0017] S1.4, based on the enterprise current working capital data, the time period corresponding to the time period enterprise receivable account data and the time period enterprise planned expenditure data, it is judged whether the sum of the enterprise current working capital data and the time period enterprise receivable account data is greater than or equal to (1+x%) of the time period enterprise planned expenditure data, and the judgment result is returned to generate the first judgment data, and x is a preset value.

[0018] Further, S2 includes the following steps:

[0019] S2.1, collecting debt object transaction repayment information, and generating debt object transaction repayment data;

[0020] S2.2, classifying the debt object transaction repayment data according to the time period, generating enterprise each time period transaction repayment data, wherein the enterprise each time period transaction repayment data of each category corresponds to the same time period:

[0021] S2.3, extracting features from the enterprise each time period transaction repayment data, and generating enterprise each time period transaction repayment feature data; wherein the features extracted during feature extraction can include historical total transaction amount, average single transaction amount, maximum / minimum single transaction amount, recent transaction amount trend, total transaction times, monthly average transaction times, average / maximum / minimum interval time, average repayment period, repayment on time rate, overdue rate, longest overdue days, account age distribution, repayment period volatility, repayment amount volatility, partial repayment frequency, etc.

[0022] S2.4, S2.4, collecting historical enterprise each time period transaction repayment feature data, and performing clustering analysis on the historical enterprise each time period transaction repayment feature data based on DBSCAN clustering algorithm, to generate a plurality of historical enterprise each time period transaction repayment feature data clusters;

[0023] S2.5, judge whether the historical enterprise each period transaction repayment characteristic data cluster is equal to the set risk index number; if yes, use the risk index to mark the historical enterprise each period transaction repayment characteristic data cluster one by one;

[0024] S2.6, if no, judge whether the number of historical enterprise each period transaction repayment characteristic data cluster is less than the set risk index number, if yes, reduce the neighborhood radius of DBSCAN clustering algorithm, return to S2.4;

[0025] S2.7, if no, calculate the center point characteristic data of each historical enterprise each period transaction repayment characteristic data cluster based on the Euclidean distance, and generate the historical enterprise each period transaction repayment center characteristic data;

[0026] Further, the target function can be set to make the sum of the Euclidean distances from the center point to each historical enterprise each period transaction repayment characteristic data in the corresponding historical enterprise each period transaction repayment characteristic data cluster minimum, and then the corresponding historical enterprise each period transaction repayment center characteristic data is obtained by solving the target function by gradient descent method.

[0027] S2.8, calculate the Euclidean distance between each historical enterprise each period transaction repayment center characteristic data and other historical enterprise each period transaction repayment center characteristic data, and merge the corresponding two historical enterprise each period transaction repayment characteristic data clusters in ascending order of the obtained Euclidean distance, to generate new historical enterprise each period transaction repayment characteristic data cluster, until the number of new historical enterprise each period transaction repayment characteristic data cluster is equal to the set risk index number;

[0028] S2.9, calculate the center point characteristic data of each new historical enterprise each period transaction repayment characteristic data cluster based on the Euclidean distance, generate new historical enterprise each period transaction repayment center characteristic data, calculate the Euclidean distance between the enterprise each period transaction repayment characteristic data and each new historical enterprise each period transaction repayment center characteristic data, select the risk index corresponding to the new historical enterprise each period transaction repayment center characteristic data corresponding to the minimum Euclidean distance, and generate the second judgment data.

[0029] Further, S3 includes the following steps:

[0030] S3.1, collecting debt object transaction repayment information to generate debt object transaction repayment data;

[0031] S3.2, based on affinity propagation clustering algorithm, clustering analysis is carried out on the debt object transaction repayment data of all historical debt objects to generate a plurality of debt object transaction repayment data clusters;

[0032] S3.3, set a plurality of object category labels corresponding to the debt object transaction repayment data cluster one by one to generate an object category label-debt object transaction repayment data cluster set;

[0033] S3.4, searching the object class label corresponding to the debt object transaction repayment data cluster matched with the debt object transaction repayment data characteristics in the object class label-debt object transaction repayment data cluster set, and generating object class label data.

[0034] Further, S3 further includes the following steps:

[0035] determining whether the debt object exists corresponding object class label data;

[0036] if yes, performing S4 step;

[0037] if no, performing S3 step.

[0038] Further, S4 includes the following steps:

[0039] S4.1, collecting other debt object transaction repayment data of the same object class label data and the debt object transaction repayment data of the current period, and generating an object class label other debt object transaction repayment data set;

[0040] S4.2, respectively extracting features from the debt object transaction repayment data of the current period and the object class label other debt object transaction repayment data set, and generating debt object transaction repayment feature data and object class label other debt object transaction repayment feature data;

[0041] S4.3, calculating and processing the differences of each feature of the debt object transaction repayment feature data and the object class label other debt object transaction repayment feature data, and generating first debt object transaction repayment feature difference data;

[0042] S4.4, based on the first debt object transaction repayment feature difference data, determining whether the debt object transaction repayment data of the current period is healthier than all other debt object transaction repayment data of the same debt object class label data, and returning the determination result to generate third determination data.

[0043] Further, S4.3 further includes the following steps:

[0044] collecting each transaction repayment feature maximum threshold data;

[0045] determining whether the absolute value of the difference data of each feature in the first debt object transaction repayment feature difference data is greater than the corresponding transaction repayment feature maximum threshold data, if the feature item greater than the corresponding transaction repayment feature maximum threshold data meets the set determination condition, then returning to S3, otherwise, performing S4.4.

[0046] The setting judgment condition has multiple, and each judgment condition is in the relationship of or, that is, one judgment condition is regarded as meeting.

[0047] Further, S5 includes the following steps:

[0048] S5.1, respectively, to the current period of debt object transaction repayment data and all periods of debt object transaction repayment data, feature extraction is carried out, and debt object transaction repayment feature data and debt object total transaction repayment feature data are generated;

[0049] S5.2, the difference of each feature of debt object transaction repayment feature data and debt object total transaction repayment feature data is calculated and processed, and second debt object transaction repayment feature difference data is generated;

[0050] S5.3, based on the second debt object transaction repayment feature difference data, whether the current period of debt object transaction repayment data is healthier than all periods of debt object transaction repayment data is judged, and the judgment result is returned to generate fourth judgment data.

[0051] Further, S6 includes the following steps:

[0052] S6.1, the first judgment data, the second judgment data, the third judgment data and the fourth judgment data are collected, and analysis feedback data are generated;

[0053] S6.2, different analysis feedback data and corresponding decision scheme data are collected, and decision scheme analysis feedback data set is generated;

[0054] S6.3, based on the naive algorithm, the decision scheme data matched with the analysis feedback data in the decision scheme analysis feedback data set is searched, the decision scheme feedback data is generated, and the analysis feedback work is executed according to the analysis feedback data and the decision scheme feedback data.

[0055] A big data analysis platform for enterprise accounts receivable recovery risk assessment, comprising a data transceiver, a storage, and a processor;

[0056] The transceiver is used for receiving and sending data.

[0057] The storage is used for storing received data and computer programs.

[0058] The processor is used for executing computer programs to realize a big data analysis method for enterprise accounts receivable recovery risk assessment.

[0059] The beneficial effects of the present application are:

[0060] The application can timely find the abnormality of the enterprise fund chain, so as to facilitate the enterprise management personnel to make processing, such as collecting overdue and near-term debt objects, or compressing the enterprise plan expenditure, so as to ensure the normality of the enterprise fund chain; the safety of the overall repayment of the enterprise is judged by analyzing the transaction and repayment information of all debt objects of the enterprise. By analyzing the health conditions of the transaction and repayment characteristics of each debt object compared with the transaction and repayment characteristics of the same type of debt object and the historical records of the debt object, the health of the individual debt object can be warned when the overall repayment of the enterprise is safe, and the debt object is warned according to the object category when the overall repayment of the enterprise is at high risk. BRIEF DESCRIPTION OF DRAWINGS

[0061] Figure 1 is a big data analysis method flow chart of the enterprise accounts receivable recovery risk assessment described in the application.

[0062] Figure 2 is a big data analysis platform schematic diagram of the enterprise accounts receivable recovery risk assessment. DETAILED DESCRIPTION

[0063] Please refer to Figure 1 A big data analysis method for enterprise accounts receivable recovery risk assessment, comprising the following steps:

[0064] S1, collecting enterprise accounts receivable data, enterprise plan expenditure data and enterprise current working capital data of each period in the future, analyzing and processing whether the working capital of the enterprise in each period in the future meets the enterprise plan expenditure data of the corresponding period, and generating first judgment data; wherein the enterprise accounts receivable data can include but is not limited to accounts receivable unique identifier, debt object unique number, associated contract number, invoice number, accounts receivable generation date, agreed payment date, pre-tax receivable amount and other data, wherein the corresponding materials can be found by number, and the corresponding detailed data can be consulted.

[0065] Specifically comprising the following steps:

[0066] S1.1, according to the time length of the accounts receivable time from the current time, a plurality of time periods are set, and a time period data set is generated; wherein the time length of each time period set according to actual needs can be selected, for example, the time period can be set as 0-3 days, 0-15 days, 0-30 days, 30-60 days, 60-90 days, 90-180 days, etc.

[0067] S1.2, collecting enterprise accounts receivable data, enterprise plan expenditure data and enterprise current working capital data;

[0068] S1.3, character matching search is performed on the period data set, enterprise receivables data and enterprise planned expenditure data, enterprise receivables data and enterprise planned expenditure data of each period are searched out, and a plurality of period enterprise receivables data and a plurality of period enterprise planned expenditure data are respectively generated;

[0069] S1.4, based on the enterprise current working capital data, the period enterprise receivables data corresponding to the period and the period enterprise planned expenditure data, a judgment process is performed on whether the sum of the enterprise current working capital data and the period enterprise receivables data is greater than or equal to (1+x%) of the period enterprise planned expenditure data, a judgment result is returned to generate first judgment data, and x is a preset value.

[0070] S2, collecting transaction repayment information of each debtor object in each period, judging whether the transaction repayment of the enterprise in each period is healthy to generate second judgment data; wherein, the transaction repayment information is divided into transaction information and repayment information, the transaction information can include transaction serial number, transaction commodity information, transaction time, transaction responsible person, etc.; the repayment information can include repayment record unique number, actual repayment date, repayment amount, delay days, repayment bank account, etc.

[0071] Specifically, the following steps are included:

[0072] S2.1, collecting transaction repayment information of the debtor object to generate transaction repayment data of the debtor object;

[0073] S2.2, classifying the transaction repayment data of the debtor object according to the period to generate transaction repayment data of the enterprise in each period, wherein the transaction repayment data of the enterprise in each period of each category corresponds to the same period:

[0074] S2.3, extracting features from the transaction repayment data of the enterprise in each period to generate transaction repayment feature data of the enterprise in each period; wherein, when the features are extracted, the extracted features can include total historical transaction amount, average single transaction amount, maximum / minimum single transaction amount, trend of recent transaction amounts, total transaction times, monthly average transaction times, average / maximum / minimum interval time, average repayment period, on-time repayment rate, overdue rate, longest overdue days, account age distribution, repayment period volatility, repayment amount volatility, partial repayment frequency, etc.

[0075] S2.4, collecting historical transaction repayment feature data of the enterprise in each period, and performing clustering analysis on the historical transaction repayment feature data of the enterprise in each period based on the DBSCAN clustering algorithm to generate a plurality of historical transaction repayment feature data clusters of the enterprise in each period;

[0076] S2.5, judge whether the historical enterprise each period transaction repayment characteristic data cluster is equal to the set risk index number; if yes, use the risk index to mark the historical enterprise each period transaction repayment characteristic data cluster one by one;

[0077] S2.6, if no, judge whether the historical enterprise each period transaction repayment characteristic data cluster number is less than the set risk index number, if yes, reduce the neighborhood radius of the DBSCAN clustering algorithm, return to S2.4;

[0078] S2.7, if no, calculate the center point characteristic data of each historical enterprise each period transaction repayment characteristic data cluster based on the Euclidean distance, and generate the historical enterprise each period transaction repayment center characteristic data;

[0079] Further, the target function can be set to make the sum of the Euclidean distances from the center point to each historical enterprise each period transaction repayment characteristic data in the corresponding historical enterprise each period transaction repayment characteristic data cluster minimum, and then the corresponding historical enterprise each period transaction repayment center characteristic data is obtained by solving the target function by gradient descent method.

[0080] S2.8, calculate the Euclidean distance between each historical enterprise each period transaction repayment center characteristic data and other historical enterprise each period transaction repayment center characteristic data, and merge the corresponding two historical enterprise each period transaction repayment characteristic data clusters in ascending order of the obtained Euclidean distance, to generate new historical enterprise each period transaction repayment characteristic data clusters, until the number of new historical enterprise each period transaction repayment characteristic data clusters is equal to the set risk index number;

[0081] Through the above steps, the historical enterprise each period transaction repayment characteristic data can be automatically clustered into several clusters by the DBSCAN clustering algorithm, so that the clustering process can discover more complex clusters to enhance the robustness to noise, and then the Euclidean distance between the cluster center points is calculated, the clusters with similar characteristics are merged, so that the new clusters are one-to-one corresponding to the preset labels.

[0082] S2.9, calculate the center point characteristic data of each new historical enterprise each period transaction repayment characteristic data cluster based on the Euclidean distance, generate new historical enterprise each period transaction repayment center characteristic data, calculate the Euclidean distance between the enterprise each period transaction repayment characteristic data and each new historical enterprise each period transaction repayment center characteristic data, select the risk index corresponding to the new historical enterprise each period transaction repayment center characteristic data with the minimum Euclidean distance, and generate the second judgment data.

[0083] By calculating the Euclidean distance between the transaction repayment feature data of the enterprise at each period and the feature center point of each new cluster, the new cluster to which the transaction repayment feature data of the enterprise at each period belongs and the corresponding preset label, i.e., the risk indicator, are determined, so that when the transaction repayment feature data of the enterprise at each period is classified, the risk of the classification result being biased towards the majority class sample due to the uneven number of samples corresponding to each risk indicator is avoided.

[0084] The risk indicator is used as a judgment basis for judging whether the transaction repayment data of the corresponding enterprise at each period is healthy, and the second judgment data is the risk indicator corresponding to the risk indicator feature data returned by the search; the risk indicator can be set as "high risk", "medium risk", "low risk" as needed, or the risk indicator can be set as "0-level risk", "1-level risk", "2-level risk"..., "n-level risk" from low to high, and n is a positive integer.

[0085] Further, before clustering and calculating the Euclidean distance, the risk indicator feature data set and the transaction repayment feature data of the enterprise at each period need to be standardized, and the standardization formula is as follows:

[0086]

[0087] wherein x i is the i-th original feature, μ is the mean of x i in all samples, and σ is the standard deviation of x i .

[0088] S3, based on the transaction repayment information of each debt object at each period, generate debt object transaction repayment data, based on the debt object transaction repayment data and all debt object historical transaction repayment data, analyze and process the debt object transaction repayment feature category corresponding to the debt object transaction repayment data, generate object category label data, including the following steps:

[0089] S3.1, collect debt object transaction repayment information, generate debt object transaction repayment data;

[0090] S3.2, based on affinity propagation clustering algorithm, cluster analysis is performed on the debt object transaction repayment data of all historical debt objects to generate a plurality of debt object transaction repayment data clusters;

[0091] S3.3, set a plurality of object category labels corresponding to the debt object transaction repayment data clusters to generate an object category label-debt object transaction repayment data cluster set;

[0092] S3.4, based on K nearest neighbor algorithm, search the object category label corresponding to the debt object transaction repayment data cluster matching the debt object transaction repayment data feature in the object category label-debt object transaction repayment data cluster set to generate object category label data.

[0093] In one embodiment, the following judgment can also be made before performing S3:

[0094] Judgment whether the debt object exists corresponding object category label data;

[0095] If yes, perform S4 step;

[0096] If no, perform S3 step.

[0097] S4, based on the current period of debt object transaction repayment data and corresponding debt object category label data corresponding to all other current period of debt object transaction repayment data, whether the debt object transaction unit repayment data is more than the corresponding debt object category label data corresponding to all other debt object transaction repayment data Healthy judgment analysis, generate third judgment data, including the following steps:

[0098] S4.1, collect other current period of debt object transaction repayment data which is the same as the object category label data and the current period of debt object transaction repayment data, generate object category label other debt object transaction repayment data set;

[0099] S4.2, respectively, on the current period of debt object transaction repayment data and object category label other debt object transaction repayment data set, feature extraction, generate debt object transaction repayment feature data and object category label other debt object transaction repayment feature data;

[0100] S4.3, the difference of each feature of debt object transaction repayment feature data and object category label other debt object transaction repayment feature data is calculated and processed, to generate the first debt object transaction repayment feature difference data;

[0101] In one embodiment, S4.3 further comprises the following steps:

[0102] Collect each transaction repayment feature maximum threshold data;

[0103] Judgment whether the absolute value of the difference data of each feature in the first debt object transaction repayment feature difference data is greater than the corresponding transaction repayment feature maximum threshold data, if the feature item greater than the corresponding transaction repayment feature maximum threshold data meets the set judgment condition, return to S3, otherwise, perform S4.4.

[0104] There are several set judgment conditions, and each judgment condition is an or relationship, that is, it is considered to be met if one of the judgment conditions is met.

[0105] For example, one of the setting judgment conditions is "the overdue rate difference is greater than the set maximum threshold data of the overdue rate, and the overdue amount difference is greater than the set maximum threshold data of the overdue amount", and the overdue rate difference and the overdue amount difference in the first debt object transaction repayment feature difference data are greater than the set maximum threshold data of the overdue rate and the set maximum threshold data of the overdue amount respectively, which indicates that the difference between the debt object transaction repayment feature and the corresponding object category is too large, and the object category label needs to be matched again for the debt object.

[0106] S4.4, based on the first debt object transaction repayment feature difference data, judging whether the transaction repayment data of the debt object in the current period is healthier than all other transaction repayment data of the debt object in the same debt object category label data corresponding to the current period, returning the judgment result to generate the third judgment data.

[0107] In one embodiment, if the overdue rate reduction amount of the transaction repayment data of the debt object in the current period is greater than or equal to the set overdue rate threshold compared with all other transaction repayment data of the debt object corresponding to the same debt object category label data in the current period, the judgment result returns yes, if the overdue rate reduction or increase amount is less than the set overdue rate threshold, the judgment result returns flat, and if the overdue rate increase amount is greater than the set overdue rate threshold, the judgment result returns no.

[0108] S5, based on the current period debt object transaction repayment data and the historical period debt object transaction repayment data, judging and analyzing whether the current period debt object transaction repayment data is healthier than the historical period debt object transaction repayment data, generating the fourth judgment data, including the following steps:

[0109] S5.1, respectively extracting the features of the current period debt object transaction repayment data and all period debt object transaction repayment data, generating the debt object transaction repayment feature data and the debt object total transaction repayment feature data;

[0110] S5.2, calculating and processing the differences of each feature of the debt object transaction repayment feature data and the debt object total transaction repayment feature data, generating the second debt object transaction repayment feature difference data;

[0111] S5.3, based on the second debt object transaction repayment feature difference data, judging whether the transaction repayment data of the debt object in the current period is healthier than the transaction repayment data of the debt object in all periods, returning the judgment result to generate the fourth judgment data.

[0112] S6, collecting the first judgment data, the second judgment data, the third judgment data and the fourth judgment data for analysis and feedback work, including the following steps:

[0113] S6.1 Collect the first judgment data, the second judgment data, the third judgment data, and the fourth judgment data, and generate analysis feedback data;

[0114] S6.2 Collect different analysis feedback data and corresponding decision scheme data, and generate a decision scheme analysis feedback data set;

[0115] S6.3 Based on a naive algorithm, search for decision scheme data that matches the analysis feedback data in the decision scheme analysis feedback data set, generate decision scheme feedback data, and perform analysis feedback operations based on the analysis feedback data and decision scheme feedback data.

[0116] For example: when the first judgment data is negative, that is, when the sum of the company's current working capital data and the company's accounts receivable data for the period is less than (1+x%) of the company's planned expenditure data for the period, the decision plan data is to suggest collecting overdue and near-due debts (the duration of accounts receivable from the current time to the set near-due threshold) and reducing the company's planned expenditures.

[0117] When the first and second judgment data are both yes, and the third and fourth judgment data are both no, the decision plan data is to suggest collecting the corresponding debt from the debtor and reducing the maximum receivables allowed for the debtor.

[0118] If the first judgment data is yes, the second judgment data is no, the third judgment data is yes, and the fourth judgment data are all no, then the company's overall accounts receivable risk is relatively high. The health of the debtor's transaction collection data is flat or declining compared to its own historical data, but its health is rising compared to the transaction collection data of other debtors with the same debtor category label during the same period. Therefore, the recommended action is to suggest pursuing collection from debtors with the same debtor category label, reducing the maximum allowable accounts receivable for the debtor, and requiring collateral guarantees.

[0119] Please see Figure 2 The present invention also provides a big data analysis platform for assessing the risk of accounts receivable collection, including a data transceiver, a memory, and a processor;

[0120] The transceiver is used for receiving and sending data;

[0121] The storage device is used to store received data and computer programs;

[0122] The processor is used to execute computer programs to implement a big data analysis method for assessing the risk of accounts receivable collection in enterprises.

Claims

1. A big data analysis method for assessing the risk of accounts receivable collection in enterprises, characterized in that, Includes the following steps: S1. Collect the company's accounts receivable data, planned expenditure data, and current working capital data for each future period. Analyze and process whether the company's working capital for each future period meets the company's planned expenditure data for the corresponding period, and generate the first judgment data. S2. Collect transaction and repayment information of each debtor at different times, judge the health of the enterprise's transaction and repayment at different times, and generate second judgment data; S3. Based on the transaction repayment information of each debt object in each time period, generate debt object transaction repayment data. Based on the debt object transaction repayment data and the historical transaction repayment data of all debt objects, analyze and process the debt object transaction repayment feature categories corresponding to the debt object transaction repayment data, and generate object category label data. S4. Based on the current period's debt object transaction repayment data and the corresponding debt object category label data, analyze whether the debt object transaction unit repayment data is healthier than the corresponding debt object category label data, and generate third judgment data. S5. Based on the current period's debt object transaction repayment data and the historical period's debt object transaction repayment data, analyze whether the current period's debt object transaction repayment data is healthier than the historical period's debt object transaction repayment data, and generate the fourth judgment data; S6. Collect the first, second, third, and fourth judgment data and perform analysis and feedback.

2. The big data analysis method for assessing the risk of accounts receivable collection as described in claim 1, characterized in that, S1 includes the following steps: S1.1 Based on the time elapsed between the accounts receivable date and the current time, set multiple time periods and generate a time period data set; S1.2 Collect enterprise accounts receivable data, enterprise planned expenditure data, and enterprise current working capital data; S1.3 Perform character matching search on the time period data set, enterprise accounts receivable data and enterprise planned expenditure data to retrieve enterprise accounts receivable data and enterprise planned expenditure data for each time period, and generate enterprise accounts receivable data and enterprise planned expenditure data for multiple time periods respectively; S1.

4. Based on the company's current working capital data, the company's accounts receivable data for the corresponding time period, and the company's planned expenditure data for the time period, determine whether the sum of the company's current working capital data and the company's accounts receivable data for the time period is greater than or equal to (1+x%) of the company's planned expenditure data for the time period, and return the judgment result to generate the first judgment data, where x is a preset value.

3. The big data analysis method for assessing the risk of accounts receivable collection as described in claim 1, characterized in that, S2 includes the following steps: S2.1 Collect transaction repayment information of debt objects and generate transaction repayment data of debt objects; S2.

2. Classify the transaction repayment data of the debtor according to the time period to generate transaction repayment data for each time period of the enterprise, wherein the time period corresponding to the transaction repayment data of each category of the enterprise is the same: S2.3 Extract features from the enterprise's transaction payment data for each time period to generate transaction payment feature data for each time period; S2.4 Collect historical enterprise transaction and payment characteristics data for each time period, and perform cluster analysis on the historical enterprise transaction and payment characteristics data for each time period based on the DBSCAN clustering algorithm to generate multiple historical enterprise transaction and payment characteristics data clusters for each time period; S2.5 Determine whether the number of historical enterprise transaction payment characteristic data clusters for each period is equal to the number of risk indicators set; if so, use risk indicators to label each historical enterprise transaction payment characteristic data cluster for each period. S2.6 If not, determine whether the number of historical enterprise transaction repayment characteristic data clusters for each period is less than the number of risk indicators set. If yes, reduce the neighborhood radius of the DBSCAN clustering algorithm and return to S2.

4. S2.7 If not, then calculate the central point feature data of the transaction repayment feature data cluster of each historical enterprise in each period based on Euclidean distance, and generate the central feature data of the transaction repayment of each historical enterprise in each period. S2.8 Calculate the Euclidean distance between the transaction return center feature data of each historical enterprise and the transaction return center feature data of other historical enterprises in each period, and merge the corresponding two historical enterprise transaction return feature data clusters in each period in ascending order of the obtained Euclidean distance to generate new historical enterprise transaction return feature data clusters in each period, until the number of new historical enterprise transaction return feature data clusters in each period is equal to the number of risk indicators set. S2.

9. Calculate the central point feature data of the transaction return feature data clusters of each new historical enterprise in each time period based on Euclidean distance, generate the central feature data of transaction return for each time period of the new historical enterprise, calculate the Euclidean distance between the transaction return feature data of each enterprise in each time period and the central feature data of transaction return for each time period of each new historical enterprise, select the risk indicator corresponding to the central feature data of transaction return for each time period of the new historical enterprise with the smallest corresponding Euclidean distance, and generate the second judgment data.

4. The big data analysis method for assessing the risk of accounts receivable collection as described in claim 1, characterized in that, S3 includes the following steps: S3.1 Collect transaction repayment information of debt objects and generate transaction repayment data of debt objects; S3.2 Based on the affinity propagation clustering algorithm, cluster analysis is performed on the debt object transaction repayment data of all historical debt objects to generate multiple debt object transaction repayment data clusters; S3.3 Set multiple object category labels to correspond one-to-one with debt object transaction repayment data clusters, and generate an object category label-debt object transaction repayment data cluster set; S3.

4. Based on the K-nearest neighbor algorithm, search for the object category label corresponding to the debt object transaction repayment data cluster that matches the features of the debt object transaction repayment data in the object category label - debt object transaction repayment data cluster set, and generate object category label data.

5. The big data analysis method for assessing the risk of accounts receivable collection as described in claim 1, characterized in that, S4 includes the following steps: S4.1 Collect other debt object transaction repayment data for the current period whose object category label data is the same as the debt object transaction repayment data for the current period, and generate a set of other debt object transaction repayment data with object category label; S4.

2. Extract features from the current period's debt object transaction repayment data and the other debt object transaction repayment data set by object category label, respectively, to generate debt object transaction repayment feature data and other debt object transaction repayment feature data by object category label; S4.3 Calculate and process the differences in various features of the debt object transaction repayment feature data and the other debt object transaction repayment feature data of the object category label to generate the first debt object transaction repayment feature difference data; S4.4 Based on the difference data of transaction repayment characteristics of the first debt object, determine whether the transaction repayment data of the debt object in the current period is healthier than the transaction repayment data of all other debt objects in the current period corresponding to the same debt object category label data, and return the judgment result to generate the third judgment data.

6. The big data analysis method for assessing the risk of accounts receivable collection as described in claim 5, characterized in that, S4.3 also includes the following steps: Collect maximum threshold data for various transaction repayment characteristics; Determine whether the absolute value of the difference data of each feature in the transaction repayment feature difference data of the first debt object is greater than the maximum threshold data of the corresponding transaction repayment feature. If the feature item that is greater than the maximum threshold data of the corresponding transaction repayment feature meets the set judgment condition, return to S3; otherwise, proceed to S4.

4. There are multiple conditions for judgment, and each condition is related to another condition by an OR relationship, meaning that satisfying any one of the conditions is considered a satisfactory judgment.

7. The big data analysis method for assessing the risk of accounts receivable collection as described in claim 1, characterized in that, S5 includes the following steps: S5.1 Extract features from the debt object transaction repayment data for the current period and the debt object transaction repayment data for all periods to generate debt object transaction repayment feature data and total debt object transaction repayment feature data. S5.2 Calculate and process the differences between the various characteristics of the debt object transaction repayment characteristic data and the total debt object transaction repayment characteristic data to generate the second debt object transaction repayment characteristic difference data; S5.3 Based on the difference data of transaction repayment characteristics of the second debt object, determine whether the transaction repayment data of the debt object in the current period is healthier than the transaction repayment data of the debt object in all periods, and return the judgment result to generate the fourth judgment data.

8. The big data analysis method for assessing the risk of accounts receivable collection as described in claim 1, characterized in that, S6 includes the following steps: S6.1 Collect the first judgment data, the second judgment data, the third judgment data, and the fourth judgment data, and generate analysis feedback data; S6.2 Collect different analysis feedback data and corresponding decision scheme data, and generate a decision scheme analysis feedback data set; S6.3 Based on a naive algorithm, search for decision scheme data that matches the analysis feedback data in the decision scheme analysis feedback data set, generate decision scheme feedback data, and perform analysis feedback operations based on the analysis feedback data and decision scheme feedback data.

9. A platform for the big data analysis method for assessing the risk of accounts receivable collection as described in claim 1, characterized in that, This includes data transceivers, memory, and processors; The transceiver is used for receiving and sending data; The storage device is used to store received data and computer programs; The processor is used to execute computer programs to implement a big data analysis method for assessing the risk of accounts receivable collection in enterprises.

Images