CN112801774A - Wind control method based on credible data behavior analysis - Google Patents
Wind control method based on credible data behavior analysis Download PDFInfo
- Publication number
- CN112801774A CN112801774A CN202110105741.6A CN202110105741A CN112801774A CN 112801774 A CN112801774 A CN 112801774A CN 202110105741 A CN202110105741 A CN 202110105741A CN 112801774 A CN112801774 A CN 112801774A
- Authority
- CN
- China
- Prior art keywords
- data
- industry
- evaluation result
- value
- months
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/03—Credit; Loans; Processing thereof
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/12—Accounting
- G06Q40/123—Tax preparation or submission
Landscapes
- Business, Economics & Management (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Engineering & Computer Science (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- Technology Law (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a wind control method based on credible data behavior analysis, which comprises the following steps: the method comprises the steps of credible data acquisition, same industry evaluation, historical behavior evaluation, future behavior prediction and comprehensive evaluation. The invention also discloses a wind control device based on the credible data behavior analysis, which comprises the following components: the system comprises a credible data acquisition module, a same-industry evaluation module, a historical behavior evaluation module, a future behavior prediction module and a comprehensive evaluation module. The wind control method and the wind control device based on the credible data behavior analysis provide a wind control method based on enterprise tax payment behavior analysis, realize enterprise risk identification and assist loan admission decision making, and can effectively solve the problems in the prior art.
Description
Technical Field
The invention relates to the field of financial science and technology, in particular to the technical field of wind control for loan of small and micro enterprises.
Background
The concept of financial pros is becoming popular at a social level, but the implementation and development of financial pros still faces a few difficulties. The difficulty mainly comes from the contradiction between the small micro-enterprise, which is the service object of the general finance, and the risk control requirement of the financial institution.
Wind control is the most interesting element of financial institutions such as banks, whose wind control architecture traditionally relies on an offline mode of operation. The off-line mode has the disadvantage of higher implementation cost, while the general finance mainly serves small and micro enterprises, the income is lower, and the income cannot cover the cost of carrying out off-line wind control. And small and miniature enterprises generally have poor risk resistance and poor operation stability, and if no wind control measures are taken, the bank is at a higher risk of providing loan.
Because the above contradiction does not have effective solving means at present, the financial service progress of the bank to the small and micro enterprises is slow. Although banks have a good mind in providing popular financial services for larger and smaller enterprises, the enterprises generally lack comprehensive credit records, and the businesses cannot achieve effective risk identification due to the lack of effective auxiliary judgment of external indexes.
In summary, the main problems facing current affordable finance are:
1) the traditional offline wind control cost is too high, and the coverage is not comprehensive;
2) and the risk is uncontrollable due to the lack of effective auxiliary judgment of external indexes.
Disclosure of Invention
The invention provides a risk control means for small and micro enterprises, and tax payment data is introduced as credible external data for risk assessment.
According to an embodiment of the invention, a wind control method based on credible data behavior analysis is provided, which includes the following steps:
a trusted data acquisition step, wherein trusted data of a target object is acquired, the trusted data is subjected to direct rejection condition screening, if the direct rejection condition is met, the target object is directly rejected, and if the direct rejection condition is not met, the trusted data is preprocessed to obtain the target data;
the method comprises the following steps of evaluating the same industry, namely determining the industry classification of a target object, acquiring the credible data of the same type of object in the industry, calculating an industry mean value and an industry discrete value based on the credible data of the same type of object in the industry, and acquiring the same industry evaluation result of the target object according to the target data of the target object, the industry mean value and the industry discrete value, wherein the same industry evaluation result is negative if the target data is worse than the industry mean value and the industry discrete value, and the same industry evaluation result is negative if the target data is worse than the industry mean value and the industry discrete;
a historical behavior evaluation step, namely dividing target data into long-term data and short-term data, and respectively performing short-period sliding window sampling and long-period sliding window sampling on the long-term data and the short-term data, wherein if the long-term data is superior to the short-term data and the dispersion of the short-period sampling is greater than that of the long-period sampling, the historical behavior evaluation result is rejected, otherwise, the historical behavior evaluation result is passed;
a future behavior prediction step of predicting a future near-term prediction value and a future far-term prediction value by using a prediction model based on target data, wherein if the future far-term prediction value is worse than the future near-term prediction value and the future near-term prediction value is worse than a current value, a future behavior evaluation result is negative, otherwise, the future behavior evaluation result is passed;
and a comprehensive evaluation step, when the same-industry evaluation result, the historical behavior evaluation result and the future behavior evaluation result are passed, the comprehensive evaluation result is passed, any one of the same-industry evaluation result, the historical behavior evaluation result or the future behavior evaluation result is rejected, and the comprehensive evaluation result is rejected.
In one embodiment, the target object is a small micro-enterprise, the trusted data is tax payment data of the enterprise, and the wind control method based on the trusted data behavior analysis is a wind control method for evaluating whether to issue a loan to the enterprise based on the tax payment behavior of the enterprise.
In one embodiment, the direct veto condition includes:
the integrity condition is continuous and complete tax payment data in the last 24 months, and enterprises with the tax payment data missing or the tax payment data duration less than 24 months in the last 24 months consider that the integrity condition is not met;
the penalty condition is tax payment penalty records in the last 24 months, and enterprises with tax payment penalty records in the last 24 months consider that the penalty condition is not met;
the information loss condition is that whether the enterprise is an information loss executor or not, and if the enterprise is the information loss executor, the enterprise does not meet the information loss condition;
if any one of the integrity condition, the punishment condition or the loss of trust condition is not met, the enterprise is rejected;
for the non-rejected enterprises, collecting tax payment data of the past 24 months from the current month, collecting the tax payment data by months and sorting the tax payment data by months as target data.
In one embodiment, in the industry evaluation step, the industry classification of the target object is determined, and the obtaining of the credible data of the same type of object in the industry comprises: and inquiring the industrial and commercial information of the enterprise to obtain an industry classification, collecting tax payment data of all the small micro enterprises in the industry classification, and carrying out normalization processing on the tax payment data of all the small micro enterprises in the industry classification by using a Max-Min rule to obtain credible data of similar objects in the industry.
In one embodiment, in the same-industry evaluation step, for credible data of similar objects in the industry, tax payment data of all small micro enterprises in the industry in the past 24 months from the current month are collected, and an average tax payment value Ai of all small micro enterprises in the industry in the past 24 months is calculated, wherein Ai is an evaluation index of an industry mean value;
for each small micro enterprise in the industry, calculating a tax receiving value An of the small micro enterprise in the past 24 months;
calculating Euclidean distances between tax payment values An and average tax payment values Ai of each small micro enterprise to n small micro enterprises participating in calculation in the industry to obtain n Euclidean distances;
calculating the mean value and the standard deviation of the n Euclidean distances to obtain the mean value Avg (n) of the Euclidean distances and the standard deviation rho (n) of the Euclidean distances, wherein the Avg (n) and the rho (n) are evaluation indexes of industry discrete values;
calculating a tax payment value Ac of the target object in the past 24 months according to the target data, and calculating a Euclidean distance Dc between the Ac and the average tax payment value Ai;
if the conditions are simultaneously satisfied: ac < Ai, and the difference between Dc and Avg (n) is more than 3 times of rho (n), the target data is considered to be worse than the industry mean value and the industry discrete value, and the evaluation result of the same industry is negative;
otherwise, the evaluation result of the same industry is passed.
In one embodiment, the tax payment data of the small micro enterprise in the past 24 months from the current month is divided into two groups, the tax payment data of the past 24-13 months is long-term data, and the tax payment data of the past 12-current month is short-term data;
and respectively sampling long-term data and short-term data in a sliding window mode by taking 3 months as a short period sliding sampling window and taking 6 months as a long period sliding sampling window.
In one embodiment, the forward date is G1, comprising 12 data points, the near date is G2, also comprising 12 data points;
performing sliding window sampling on the long-term data G1 by taking 3 months as a short-period sliding sampling window, wherein the monthly average value of the sampling is taken as a short-period sampling average value S1 of the long-term data, and the long-term data comprises 10 data points;
taking 3 months as a short-period sliding sampling window to perform sliding window sampling on the recent data G2, wherein the monthly average value of the sampling is taken as a short-period sampling average value S2 of the recent data, and the recent data G2 comprises 10 data points;
taking 6 months as a long-period sliding sampling window to perform sliding window sampling on the long-period data G1, wherein the monthly average value of the sampling is taken as a long-period sampling average value H1 of the long-period data, and the long-period sampling average value H1 comprises 7 data points;
taking 6 months as a long period sliding sampling window to perform sliding window sampling on the recent data G2, wherein the monthly average value of the sampling is taken as a long period sampling average value H2 of the recent data and comprises 7 data points;
calculating Euclidean distance between data points corresponding to S1 and S2 to obtain a mean value D (S) of the Euclidean distance, wherein D (S) is the dispersion of short-period sampling, calculating the Euclidean distance between the data points corresponding to H1 and H2 to obtain a mean value D (H) of the Euclidean distance, and D (H) is the dispersion of long-period sampling;
calculating a mean Avg of S1 (S1), a mean Avg of S2 (S2), a mean Avg of H1 (H1), and a mean Avg of H2 (H2);
if the conditions are simultaneously satisfied: when Avg (S1) > Avg (S2), Avg (H1) > Avg (H2) and D (S) > D (H), the long-term data is better than the short-term data, the dispersion of the short-period samples is greater than that of the long-period samples, and the historical behavior evaluation result is negative;
otherwise, the historical behavior evaluation result is passed.
In one embodiment, a prophet predictive model is used to predict future near term predicted values and future predicted values based on target data, namely tax data of the small micro-enterprise for the past 24 months from the current month;
establishing a prophet prediction model according to the data window of 6 months and the target window of 3 months, and outputting a predicted value of 3 months in the future as a future near-term predicted value, which is marked as F (3);
establishing a prophet prediction model according to the data window of 12 months and the target window of 6 months, outputting a predicted value of the future 6 months as a future long-term predicted value, and marking as F (6);
if the condition is satisfied: f (6) < F (3) < tax acceptance value of current month, the future behavior evaluation result is negative, otherwise the future behavior evaluation result is passed.
According to an embodiment of the present invention, a wind control device based on behavior analysis of trusted data is provided, including: the system comprises a credible data acquisition module, a same-industry evaluation module, a historical behavior evaluation module, a future behavior prediction module and a comprehensive evaluation module. The trusted data acquisition module acquires trusted data of a target object, directly rejects the trusted data for screening, directly rejects the target object if the trusted data meets a direct rejection condition, and preprocesses the trusted data if the trusted data does not meet the direct rejection condition to obtain the target data. The same-industry evaluation module determines the industry classification of the target object, obtains the credible data of the same-industry object in the industry, calculates an industry mean value and an industry discrete value based on the credible data of the same-industry object in the industry, obtains the same-industry evaluation result of the target object according to the target data of the target object, the industry mean value and the industry discrete value, and if the target data is worse than the industry mean value and the industry discrete value, the same-industry evaluation result is negative, otherwise, the same-industry evaluation result is passed. The historical behavior evaluation module divides the target data into long-term data and short-term data, short-period sliding window sampling and long-period sliding window sampling are respectively carried out on the long-term data and the short-term data, if the long-term data are superior to the short-term data, and the dispersion of the short-period sampling is greater than that of the long-period sampling, the historical behavior evaluation result is rejected, otherwise, the historical behavior evaluation result is passed. The future behavior prediction module predicts a future near predicted value and a future predicted value by using a prediction model based on the target data, and if the future predicted value is worse than the future near predicted value and the future near predicted value is worse than a current value, the future behavior evaluation result is negative, otherwise, the future behavior evaluation result is passed. When the same-industry evaluation result, the historical behavior evaluation result and the future behavior evaluation result are passed, the comprehensive evaluation result of the comprehensive evaluation module is passed, any one of the same-industry evaluation result, the historical behavior evaluation result or the future behavior evaluation result is rejected, and the comprehensive evaluation result of the comprehensive evaluation module is rejected.
The wind control method and the wind control device based on the credible data behavior analysis provide a wind control method based on enterprise tax payment behavior analysis, realize enterprise risk identification and assist loan admission decision making. The tax payment data of the enterprise is provided by the tax bureau, and has natural higher credibility and reference value. Through the interactive platform of butt joint silver tax, acquire the tax data of enterprise to processing for external index, carrying out the supplementary judgement of loan admission through the online mode, can solving two problems that exist among the prior art: the tax payment data of enterprises are collected for analysis, and the collected tax payment data is used as an effective external index for filtering and screening target customers to carry out risk control. The online release and the automatic approval are carried out, the problem of high cost under the traditional line is solved, and the service coverage rate of the general finance is improved.
Drawings
Fig. 1 discloses a flow chart of a wind control method based on trusted data behavior analysis according to an embodiment of the present invention.
Fig. 2 discloses a functional block diagram of a wind control device based on trusted data behavior analysis according to an embodiment of the present invention.
Detailed Description
Fig. 1 discloses a flow chart of a wind control method based on trusted data behavior analysis according to an embodiment of the present invention. In one embodiment, the target object of the application of the wind control method is a small and miniature enterprise, the adopted credible data is tax payment data of the enterprise, and the wind control method based on the credible data behavior analysis is a wind control method for evaluating whether to issue a loan to the enterprise based on the tax payment behavior of the enterprise. Referring to fig. 1, the wind control method based on the credible data behavior analysis includes the following steps:
and S1, acquiring the credible data. And acquiring trusted data of the target object, screening the trusted data under a direct rejection condition, directly rejecting the target object if the direct rejection condition is met, and preprocessing the trusted data if the direct rejection condition is not met to obtain the target data. In one embodiment, the direct veto condition includes: integrity conditions, penalty conditions, and loss of trust conditions.
The integrity condition is continuous and complete tax payment data in the last 24 months, and enterprises with missing tax payment data in the last 24 months or with the duration of the tax payment data being less than 24 months consider that the integrity condition is not met.
The penalty condition is tax penalty records in the last 24 months, and enterprises with tax penalty records in the last 24 months consider that the penalty condition is not met.
The information loss condition is whether the enterprise is an information loss executor or not, and if the enterprise is the information loss executor, the enterprise does not meet the information loss condition.
If any of the integrity condition, the penalty condition, or the loss of trust condition is not met, the enterprise is directly denied.
For the non-rejected enterprises, collecting tax payment data of the past 24 months from the current month, collecting the tax payment data by months and sorting the tax payment data by months as target data. There are typically 24 data points in the target data, each of which is a tax value for a month corresponding to the business.
S2, performing industry evaluation, namely determining the industry classification of the target object, acquiring the credible data of the same type of object in the industry, calculating an industry mean value and an industry discrete value based on the credible data of the same type of object in the industry, and acquiring the industry evaluation result of the target object according to the target data of the target object, the industry mean value and the industry discrete value, wherein the industry evaluation result is negative if the target data is worse than the industry mean value and the industry discrete value, and otherwise, the industry evaluation result is passed.
In one embodiment, in the industry assessment step S2, determining an industry classification of the target object, and obtaining the credible data of the same kind of object in the industry includes: and inquiring the industrial and commercial information of the enterprise to obtain an industry classification, collecting tax payment data of all the small micro enterprises in the industry classification, and carrying out normalization processing on the tax payment data of all the small micro enterprises in the industry classification by using a Max-Min rule to obtain credible data of similar objects in the industry.
Specifically, in one embodiment, in the same-industry evaluation step, for the credible data of the same-class object (i.e. the same small micro enterprise) in the industry which has undergone the normalization process, tax payment data of all the small micro enterprises in the industry in the past 24 months from the current month is collected, and the average tax payment value Ai of all the small micro enterprises in the industry in the past 24 months is calculated, and Ai is an evaluation index of the industry average. The specific calculation method of Ai is as follows: and for each month, calculating the total amount of the tax payment data of all the small and miniature enterprises in the industry of the month, dividing the total amount of the small and miniature enterprises which are continuously stored in the month to obtain the tax payment average value of the small and miniature enterprises in the industry of the month, and accumulating the data of 24 months to obtain the average tax payment value Ai of all the small and miniature enterprises in the industry of 24 months. Because the small and miniature enterprises have weak operation stability and short duration, the calculation is more accurate in a mode of respectively calculating and accumulating every month.
For each small micro enterprise in the industry, calculate the revenue value An of the small micro enterprise in the past 24 months. If there are a total of n small micro-businesses, then n data points An will be obtained.
And calculating Euclidean distances between the tax payment value An and the average tax payment value Ai of each small micro enterprise to obtain n Euclidean distances for the n small micro enterprises in the industry participating in calculation.
And calculating the mean value and the standard deviation of the n Euclidean distances to obtain the mean value Avg (n) of the Euclidean distances and the standard deviation rho (n) of the Euclidean distances, wherein the Avg (n) and the rho (n) are evaluation indexes of industry discrete values.
According to the target data, calculating the tax payment value Ac of the target object, namely the small micro enterprise applying for loan in the last 24 months, wherein the tax payment value Ac is the total tax payment amount of the small micro enterprise in the last 24 months, and then calculating the Euclidean distance Dc between Ac and the average tax payment value Ai.
If the conditions are simultaneously satisfied: ac < Ai, and the difference between Dc and Avg (n) is more than 3 times of rho (n), the target data is considered to be worse than the industry mean value and the industry discrete value, the tax data and the business quality of the enterprise are lower than the industry average level, and the evaluation result of the same industry is rejected. If the above conditions are not simultaneously satisfied, the result of the evaluation by the same industry is passed.
And S3, historical behavior evaluation, namely dividing the target data into long-term data and short-term data, respectively carrying out short-period sliding window sampling and long-period sliding window sampling on the long-term data and the short-term data, if the long-term data is better than the short-term data and the dispersion of the short-period sampling is greater than that of the long-period sampling, the historical behavior evaluation result is rejected, otherwise, the historical behavior evaluation result is passed.
In one embodiment, the tax payment data of the small micro enterprise in the past 24 months from the current month is divided into two groups, the tax payment data of the past 24-13 months is long-term data, and the tax payment data of the past 12-current month is short-term data. And respectively sampling long-term data and short-term data in a sliding window mode by taking 3 months as a short period sliding sampling window and taking 6 months as a long period sliding sampling window.
Specifically, the forward date data is G1 including 12 data points in the past 24 to 13 months (one data point per month), and the near date data is G2 including 12 data points in the past 12 to the current month.
The long-term data G1 is sliding-window sampled with a short-period sliding sampling window of 3 months, and the monthly average of the samples (i.e., the 3-month monthly average in the window period) is used as the short-period sampling average S1 of the long-term data, which includes 10 data points (10 window sampling periods).
The recent data G2 is sampled by sliding window with 3 months as short period sliding sampling window, and the monthly average value of the sampling (i.e. the monthly average value of 3 months in the window period) is taken as the short period sampling average value S2 of the recent data, which includes 10 data points (10 window sampling periods).
The long-term data G1 is sliding-window sampled with a long-period sliding sampling window of 6 months, and the average value of the sampled months (i.e., the average value of 6 months in the window period) is used as the long-period sampling average value H1 of the long-term data, which includes 7 data points (7 window sampling periods).
The recent data G2 is sampled by sliding window with 6 months as long period sliding sampling window, and the monthly average value of the sampling (i.e. the monthly average value of 6 months in the window period) is taken as the long period sampling average value H2 of the recent data, which includes 7 data points (7 window sampling periods).
Calculating Euclidean distances between data points corresponding to S1 and S2, arranging 10 data points of S1 in sequence, arranging 10 data points of S2 in sequence, calculating the Euclidean distances between data points with corresponding orders, and obtaining a mean value D (S) of the Euclidean distances, wherein D (S) is the dispersion of short-period sampling. And calculating the Euclidean distance between the data points corresponding to H1 and H2, wherein 7 data points of H1 are arranged in sequence, 7 data points of H2 are also arranged in sequence, calculating the Euclidean distance between the data points with corresponding orders, and obtaining the mean value D (H) of the Euclidean distances, wherein D (H) is the dispersion of long-period sampling.
The mean Avg of S1 (S1), the mean Avg of S2 (S2), the mean Avg of H1 (H1), and the mean Avg of H2 (H2) were calculated.
If the conditions are simultaneously satisfied: when Avg (S1) > Avg (S2), Avg (H1) > Avg (H2) and D (S) > D (H), it shows that the long-term data is better than the short-term data, and the dispersion of the short-period samples is larger than that of the long-period samples, which shows that the tax payment data of the enterprise is gradually deteriorated, the fluctuation is increased, the operation stability is reduced, and the evaluation result of the historical behavior is negative. If the above conditions are not simultaneously satisfied, the historical behavior evaluation result is a pass.
And S4, a future behavior prediction step, wherein a prediction model is used for predicting a future near predicted value and a future distant predicted value based on the target data, if the future distant predicted value is worse than the future near predicted value and the future near predicted value is worse than the current value, the future behavior evaluation result is negative, otherwise, the future behavior evaluation result is passed.
In one embodiment, a prophet predictive model is used to predict future near term and future term forecasts based on target data, i.e., tax data for the small micro-enterprise for the past 24 months from the current month. prophet is an open source python prediction library developed by Facebook, is a time series prediction model, and mainly models according to time and values to predict values in a future period of time. For the introduction and code of prophetts, reference may be made to https:// github.
In one embodiment, a prophet prediction model is established according to a data window of 6 months and a target window of 3 months, and a predicted value of the future 3 months is output as a future near-term predicted value, which is denoted as F (3).
And establishing a prophet prediction model according to the data window of 12 months and the target window of 6 months, and outputting a predicted value of the future 6 months as a future long-term predicted value, which is marked as F (6).
If the condition is satisfied: f (6) < F (3) < tax value in the current month, which shows that the future tax payment data and the operation quality of the enterprise are reduced, and the future behavior evaluation result is negative. If the above condition is not satisfied, the future behavior evaluation result is a pass.
And S5, a comprehensive evaluation step, wherein when the same industry evaluation result, the historical behavior evaluation result and the future behavior evaluation result are passed, the comprehensive evaluation result is passed, any one of the same industry evaluation result, the historical behavior evaluation result or the future behavior evaluation result is rejected, and the comprehensive evaluation result is rejected.
Fig. 2 discloses a functional block diagram of a wind control device based on trusted data behavior analysis according to an embodiment of the present invention. Referring to fig. 2, the wind control device based on the trusted data behavior analysis includes: the system comprises a credible data acquisition module 101, a same-industry evaluation module 102, a historical behavior evaluation module 103, a future behavior prediction module 104 and a comprehensive evaluation module 105.
The trusted data acquisition module 101 acquires trusted data of a target object, performs direct rejection condition screening on the trusted data, directly rejects the target object if a direct rejection condition is met, and preprocesses the trusted data if a direct rejection condition is not met to obtain the target data. The trusted data collection module 101 performs the function corresponding to the aforementioned step S1.
The industry evaluation module 102 determines an industry classification of the target object, obtains credible data of the same type of object in the industry, calculates an industry mean value and an industry discrete value based on the credible data of the same type of object in the industry, obtains an industry evaluation result of the target object according to the target data of the target object, the industry mean value and the industry discrete value, and if the target data is worse than the industry mean value and the industry discrete value, the industry evaluation result is negative, otherwise, the industry evaluation result is passed. The function performed by the industry evaluation module 102 corresponds to step S2 described above.
The historical behavior evaluation module 103 divides the target data into long-term data and short-term data, and performs short-period sliding window sampling and long-period sliding window sampling on the long-term data and the short-term data respectively, if the long-term data is better than the short-term data and the dispersion of the short-period sampling is greater than that of the long-period sampling, the historical behavior evaluation result is rejected, otherwise, the historical behavior evaluation result is passed. The function performed by the historical behavior evaluation module 103 corresponds to the aforementioned step S3.
The future behavior prediction module 104 predicts a future near future predicted value and a future predicted value using a prediction model based on the target data, and if the future predicted value is worse than the future near future predicted value and the future near future predicted value is worse than the current value, the future behavior evaluation result is negative, otherwise the future behavior evaluation result is positive. The function performed by the future behavior prediction module 104 corresponds to the aforementioned step S4.
And when the same-industry evaluation result, the historical behavior evaluation result and the future behavior evaluation result are passed, the comprehensive evaluation module 105 determines that the comprehensive evaluation result is passed, determines that any one of the same-industry evaluation result, the historical behavior evaluation result or the future behavior evaluation result is negative, and determines that the comprehensive evaluation result is negative. The function performed by the comprehensive evaluation module 105 corresponds to the aforementioned step S5.
The wind control method and the wind control device based on the credible data behavior analysis provide a wind control method based on enterprise tax payment behavior analysis, realize enterprise risk identification and assist loan admission decision making. The tax payment data of the enterprise is provided by the tax bureau, and has natural higher credibility and reference value. Through the interactive platform of butt joint silver tax, acquire the tax data of enterprise to processing for external index, carrying out the supplementary judgement of loan admission through the online mode, can solving two problems that exist among the prior art: the tax payment data of enterprises are collected for analysis, and the collected tax payment data is used as an effective external index for filtering and screening target customers to carry out risk control. The online release and the automatic approval are carried out, the problem of high cost under the traditional line is solved, and the service coverage rate of the general finance is improved.
It should also be noted that the above-mentioned embodiments are only specific embodiments of the present invention. It is apparent that the present invention is not limited to the above embodiments and similar changes or modifications can be easily made by those skilled in the art from the disclosure of the present invention and shall fall within the scope of the present invention. The embodiments described above are provided to enable persons skilled in the art to make or use the invention and that modifications or variations can be made to the embodiments described above by persons skilled in the art without departing from the inventive concept of the present invention, so that the scope of protection of the present invention is not limited by the embodiments described above but should be accorded the widest scope consistent with the innovative features set forth in the claims.
Claims (9)
1. A wind control method based on credible data behavior analysis is characterized by comprising the following steps:
a trusted data acquisition step, wherein trusted data of a target object is acquired, the trusted data is subjected to direct rejection condition screening, if the direct rejection condition is met, the target object is directly rejected, and if the direct rejection condition is not met, the trusted data is preprocessed to obtain the target data;
the method comprises the following steps of evaluating the same industry, namely determining the industry classification of a target object, acquiring the credible data of the same type of object in the industry, calculating an industry mean value and an industry discrete value based on the credible data of the same type of object in the industry, and acquiring the same industry evaluation result of the target object according to the target data of the target object, the industry mean value and the industry discrete value, wherein the same industry evaluation result is negative if the target data is worse than the industry mean value and the industry discrete value, and the same industry evaluation result is negative if the target data is worse than the industry mean value and the industry discrete;
a historical behavior evaluation step, namely dividing target data into long-term data and short-term data, and respectively performing short-period sliding window sampling and long-period sliding window sampling on the long-term data and the short-term data, wherein if the long-term data is superior to the short-term data and the dispersion of the short-period sampling is greater than that of the long-period sampling, the historical behavior evaluation result is rejected, otherwise, the historical behavior evaluation result is passed;
a future behavior prediction step of predicting a future near-term prediction value and a future far-term prediction value by using a prediction model based on target data, wherein if the future far-term prediction value is worse than the future near-term prediction value and the future near-term prediction value is worse than a current value, a future behavior evaluation result is negative, otherwise, the future behavior evaluation result is passed;
and a comprehensive evaluation step, when the same-industry evaluation result, the historical behavior evaluation result and the future behavior evaluation result are passed, the comprehensive evaluation result is passed, any one of the same-industry evaluation result, the historical behavior evaluation result or the future behavior evaluation result is rejected, and the comprehensive evaluation result is rejected.
2. The wind control method based on the credible data behavior analysis according to claim 1, wherein the target object is a small micro-enterprise, the credible data is tax payment data of the enterprise, and the wind control method based on the credible data behavior analysis is a wind control method for evaluating whether to issue a loan to the enterprise based on the tax payment behavior of the enterprise.
3. The wind control method based on trusted data behavior analysis according to claim 2, wherein the direct veto condition comprises:
the integrity condition is continuous and complete tax payment data in the last 24 months, and enterprises with the tax payment data missing or the tax payment data duration less than 24 months in the last 24 months consider that the integrity condition is not met;
the penalty condition is tax payment penalty records in the last 24 months, and enterprises with tax payment penalty records in the last 24 months consider that the penalty condition is not met;
the information loss condition is that whether the enterprise is an information loss executor or not, and if the enterprise is the information loss executor, the enterprise does not meet the information loss condition;
if any one of the integrity condition, the punishment condition or the loss of trust condition is not met, the enterprise is rejected;
for the non-rejected enterprises, collecting tax payment data of the past 24 months from the current month, collecting the tax payment data by months and sorting the tax payment data by months as target data.
4. The wind control method based on credible data behavior analysis as claimed in claim 3, wherein in the same industry evaluation step, determining industry classification of the target object, and acquiring credible data of the same kind of object in the industry comprises: and inquiring the industrial and commercial information of the enterprise to obtain an industry classification, collecting tax payment data of all the small micro enterprises in the industry classification, and carrying out normalization processing on the tax payment data of all the small micro enterprises in the industry classification by using a Max-Min rule to obtain credible data of similar objects in the industry.
5. The wind control method based on credible data behavior analysis as claimed in claim 4, wherein in the same industry evaluation step, for credible data of the same kind of objects in the industry, tax payment data of all small micro-enterprises in the industry in the past 24 months from the current month is collected, and the average tax payment value Ai of all small micro-enterprises in the industry in the past 24 months is calculated, wherein Ai is an evaluation index of the industry average;
for each small micro enterprise in the industry, calculating a tax receiving value An of the small micro enterprise in the past 24 months;
calculating Euclidean distances between tax payment values An and average tax payment values Ai of each small micro enterprise to n small micro enterprises participating in calculation in the industry to obtain n Euclidean distances;
calculating the mean value and the standard deviation of the n Euclidean distances to obtain the mean value Avg (n) of the Euclidean distances and the standard deviation rho (n) of the Euclidean distances, wherein the Avg (n) and the rho (n) are evaluation indexes of industry discrete values;
calculating a tax payment value Ac of the target object in the past 24 months according to the target data, and calculating a Euclidean distance Dc between the Ac and the average tax payment value Ai;
if the conditions are simultaneously satisfied: ac < Ai, and the difference between Dc and Avg (n) is more than 3 times of rho (n), the target data is considered to be worse than the industry mean value and the industry discrete value, and the evaluation result of the same industry is negative;
otherwise, the evaluation result of the same industry is passed.
6. The wind control method based on credible data behavior analysis as claimed in claim 3, wherein the tax payment data of the small micro enterprise in the past 24 months from the current month is divided into two groups, the tax payment data of the past 24-13 months is long term data, and the tax payment data of the past 12-current month is short term data;
and respectively sampling long-term data and short-term data in a sliding window mode by taking 3 months as a short period sliding sampling window and taking 6 months as a long period sliding sampling window.
7. The wind control method based on credible data behavior analysis as claimed in claim 6, wherein the forward data is G1 and comprises 12 data points, and the near data is G2 and also comprises 12 data points;
performing sliding window sampling on the long-term data G1 by taking 3 months as a short-period sliding sampling window, wherein the monthly average value of the sampling is taken as a short-period sampling average value S1 of the long-term data, and the long-term data comprises 10 data points;
taking 3 months as a short-period sliding sampling window to perform sliding window sampling on the recent data G2, wherein the monthly average value of the sampling is taken as a short-period sampling average value S2 of the recent data, and the recent data G2 comprises 10 data points;
taking 6 months as a long-period sliding sampling window to perform sliding window sampling on the long-period data G1, wherein the monthly average value of the sampling is taken as a long-period sampling average value H1 of the long-period data, and the long-period sampling average value H1 comprises 7 data points;
taking 6 months as a long period sliding sampling window to perform sliding window sampling on the recent data G2, wherein the monthly average value of the sampling is taken as a long period sampling average value H2 of the recent data and comprises 7 data points;
calculating Euclidean distance between data points corresponding to S1 and S2 to obtain a mean value D (S) of the Euclidean distance, wherein D (S) is the dispersion of short-period sampling, calculating the Euclidean distance between the data points corresponding to H1 and H2 to obtain a mean value D (H) of the Euclidean distance, and D (H) is the dispersion of long-period sampling;
calculating a mean Avg of S1 (S1), a mean Avg of S2 (S2), a mean Avg of H1 (H1), and a mean Avg of H2 (H2);
if the conditions are simultaneously satisfied: when Avg (S1) > Avg (S2), Avg (H1) > Avg (H2) and D (S) > D (H), the long-term data is better than the short-term data, the dispersion of the short-period samples is greater than that of the long-period samples, and the historical behavior evaluation result is negative;
otherwise, the historical behavior evaluation result is passed.
8. The wind control method based on credible data behavior analysis as claimed in claim 3, wherein a prophet prediction model is used to predict future near term prediction value and future long term prediction value based on target data, namely tax payment data of the small micro-enterprise for the past 24 months from the current month;
establishing a prophet prediction model according to the data window of 6 months and the target window of 3 months, and outputting a predicted value of 3 months in the future as a future near-term predicted value, which is marked as F (3);
establishing a prophet prediction model according to the data window of 12 months and the target window of 6 months, outputting a predicted value of the future 6 months as a future long-term predicted value, and marking as F (6);
if the condition is satisfied: f (6) < F (3) < tax acceptance value of current month, the future behavior evaluation result is negative, otherwise the future behavior evaluation result is passed.
9. A wind control device based on credible data behavior analysis is characterized by comprising:
the trusted data acquisition module is used for acquiring trusted data of the target object, screening the trusted data under a direct rejection condition, directly rejecting the target object if the direct rejection condition is met, and preprocessing the trusted data if the direct rejection condition is not met to obtain the target data;
the same-industry evaluation module is used for determining the industry classification of the target object, acquiring the credible data of the same-type object in the industry, calculating an industry mean value and an industry discrete value based on the credible data of the same-type object in the industry, and acquiring the same-industry evaluation result of the target object according to the target data of the target object, the industry mean value and the industry discrete value, wherein the same-industry evaluation result is negative if the target data is worse than the industry mean value and the industry discrete value, and the same-industry evaluation result is negative if the target data is worse than the industry mean value and the industry discrete value;
the historical behavior evaluation module is used for dividing the target data into long-term data and short-term data, respectively carrying out short-period sliding window sampling and long-period sliding window sampling on the long-term data and the short-term data, if the long-term data is better than the short-term data and the dispersion of the short-period sampling is greater than that of the long-period sampling, the historical behavior evaluation result is negative, otherwise, the historical behavior evaluation result is passed;
the future behavior prediction module is used for predicting a future near-term prediction value and a future far-term prediction value by using a prediction model based on target data, if the future far-term prediction value is worse than the future near-term prediction value and the future near-term prediction value is worse than a current value, the future behavior evaluation result is negative, otherwise, the future behavior evaluation result is passed;
and when the same-industry evaluation result, the historical behavior evaluation result and the future behavior evaluation result are passed, the comprehensive evaluation result is passed, any one of the same-industry evaluation result, the historical behavior evaluation result or the future behavior evaluation result is rejected, and the comprehensive evaluation result is rejected.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110105741.6A CN112801774A (en) | 2021-01-26 | 2021-01-26 | Wind control method based on credible data behavior analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110105741.6A CN112801774A (en) | 2021-01-26 | 2021-01-26 | Wind control method based on credible data behavior analysis |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112801774A true CN112801774A (en) | 2021-05-14 |
Family
ID=75811940
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110105741.6A Pending CN112801774A (en) | 2021-01-26 | 2021-01-26 | Wind control method based on credible data behavior analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112801774A (en) |
-
2021
- 2021-01-26 CN CN202110105741.6A patent/CN112801774A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| KR102009309B1 (en) | Management automation system for financial products and management automation method using the same | |
| CN110930038A (en) | Loan demand identification method, loan demand identification device, loan demand identification terminal and loan demand identification storage medium | |
| CN112598500A (en) | Credit processing method and system for non-limit client | |
| CN109086954B (en) | Prediction method, device, equipment and medium for predicting yield based on fund flow | |
| CN112801774A (en) | Wind control method based on credible data behavior analysis | |
| CN112132679A (en) | Product recommendation method and device | |
| CN115907837A (en) | Futures data analysis and risk prediction method and system based on machine learning | |
| CN115860924A (en) | Supply chain financial credit risk early warning method and related equipment | |
| CN115905655A (en) | User portrait construction method, device and equipment and readable storage medium | |
| CN115660803A (en) | Supply chain financial risk management method under internet background | |
| CN115099933A (en) | Service budget method, device and equipment | |
| CN115049137A (en) | Prediction method and device of transaction yield, storage medium and electronic equipment | |
| CN110288482B (en) | Steel mill futures trading method and system | |
| CN112884301A (en) | Method, equipment and computer storage medium for enterprise risk analysis | |
| CN111311418A (en) | Foreign exchange rate prediction method based on combination of neural network model and classification model | |
| Chandrika et al. | Comparative analysis of machine learning algorithms to forecast Indian stock market | |
| Komala et al. | Pre indication of Stock Price Using Prolonged Petite-Stretch Memory (PPSM) Algorithm | |
| KR102566466B1 (en) | Alternative Credit Rating System for Evaluating Personal Credit | |
| Jorgenson et al. | XST XSTATE RESOURCES LIMITED | |
| Weitzman et al. | IBCP Independent Bank Corporation Common Stock | |
| Yen | New investment strategy based on the big data-volatility index breaking out signal rule | |
| Miyai et al. | Environmental, Social, and Governance Investment and Material Disclosure by Companies | |
| Eriksson et al. | Developing a selection of credit scoring models based on customer data | |
| CN115907825A (en) | Financial product exit early warning method and system, electronic equipment and storage medium | |
| Jorgenson et al. | KAU Stock Forecast: A Buy For The Next 4 Weeks |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |