CN112700321A - Multi-rule anti-fraud prediction method and system based on user behavior data - Google Patents
Multi-rule anti-fraud prediction method and system based on user behavior data Download PDFInfo
- Publication number
- CN112700321A CN112700321A CN202011631672.4A CN202011631672A CN112700321A CN 112700321 A CN112700321 A CN 112700321A CN 202011631672 A CN202011631672 A CN 202011631672A CN 112700321 A CN112700321 A CN 112700321A
- Authority
- CN
- China
- Prior art keywords
- information
- repayment
- fraud
- fraud risk
- client
- 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
- 238000000034 method Methods 0.000 title claims abstract description 30
- 238000012502 risk assessment Methods 0.000 claims abstract description 46
- 238000011156 evaluation Methods 0.000 claims abstract description 35
- 230000003111 delayed effect Effects 0.000 claims description 53
- 230000006399 behavior Effects 0.000 claims description 43
- 238000012216 screening Methods 0.000 claims description 9
- 238000012549 training Methods 0.000 claims description 7
- 230000002265 prevention Effects 0.000 claims description 5
- 238000011835 investigation Methods 0.000 claims description 4
- 238000012217 deletion Methods 0.000 description 2
- 230000037430 deletion Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000012797 qualification Methods 0.000 description 2
- 241000209202 Bromus secalinus Species 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000012790 confirmation Methods 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Software Systems (AREA)
- General Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Data Mining & Analysis (AREA)
- Strategic Management (AREA)
- Marketing (AREA)
- Economics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Technology Law (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Development Economics (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to the technical field of credit security, and particularly discloses a multi-rule anti-fraud prediction system based on user behavior data and a method of an application system. The system comprises a client information acquisition module, a fraud risk evaluation module and an anti-fraud prediction module; the fraud risk evaluation module is used for carrying out fraud risk analysis on the obtained client information through a fraud risk analysis model to obtain a fraud risk evaluation report of the client, wherein the fraud risk evaluation report comprises the client information of the existing fraud risk and corresponding fraud risk analysis information; and the anti-fraud prediction module is used for calculating the fraud risk score of the whole customer according to the preset fraud risk weight proportion rule and the customer information of the fraud risk in the fraud risk evaluation report. By adopting the technical scheme of the invention, the fraud risk of the customer can be more accurately analyzed, so that the loss of the bank caused by the bad loan is reduced.
Description
Technical Field
The invention relates to the technical field of credit security, in particular to a multi-rule anti-fraud prediction method and a multi-rule anti-fraud prediction system based on user behavior data.
Background
At present, with the rise of the internet finance, the loan is more convenient and faster, and meanwhile, the phenomena of credit risk and user fraud are more obvious. Wherein:
the credit risk is also called default risk, which means the possibility that a borrower, a security issuer or a transaction counterpart will lose money due to unwilling or inability to fulfill contract conditions for various reasons, thereby causing loss to banks, investors or transaction partners. Credit risk is a major risk that banks face serious risk problems if they cannot identify lost assets in time, increase the reserve to be credited and stop interest income confirmation under appropriate conditions.
User fraud refers to fraud in loans at banks or other financial institutions for the purpose of illicit occupations, creating false reasons for introducing funds, projects, etc., using false economic contracts, documenting documents, or using false proof-of-property warranties and other methods where false facts hide true facts.
Refers to the risk that the counterparty does not fulfill the debt due.
With the rapid expansion of the credit market scale in developing countries, the development prospects of banks are also changing. Meanwhile, how to prevent personal credit fraud and reduce bad credit rate has become an important research topic of commercial banks. Although the scale of the domestic credit market has been greatly increased, the quality of the corresponding service has not been significantly improved, and particularly, commercial banks face various problems, such as overdue loan and bad loan, which cause significant loss to the commercial banks.
Disclosure of Invention
In order to solve the technical problem of accurately analyzing the fraud risk of a customer so as to reduce the loss of a bank caused by bad loan, the invention provides a multi-rule anti-fraud prediction method and a multi-rule anti-fraud prediction system based on user behavior data.
The basic scheme of the invention is as follows:
the multi-rule anti-fraud prediction system based on user behavior data comprises a client information acquisition module, a fraud risk evaluation module and an anti-fraud prediction module, wherein:
the client information acquisition module is used for acquiring client information, wherein the client information comprises historical transaction behaviors, personal credit investigation, personal condition information, identity information and three-party data;
the fraud risk evaluation module is used for carrying out fraud risk analysis on the obtained client information through a fraud risk analysis model to obtain a fraud risk evaluation report of the client, wherein the fraud risk evaluation report comprises the client information of the existing fraud risk and corresponding fraud risk analysis information;
and the fraud prevention prediction module is used for calculating the fraud risk score of the whole customer according to the preset fraud risk weight proportion rule and the customer information of the fraud risk in the fraud risk evaluation report, wherein the fraud risk weight proportion rule is a weight proportion set according to the influence importance degree of each information in the customer information on the fraud risk.
The working principle and the advantages of the scheme are as follows:
1. the client information acquisition module is used for acquiring client information (identity information, personal condition information, personal credit and historical transaction behaviors) through direct (directly provided by the client) and indirect (a third-party credit investigation platform and a third-party data platform) so as to support data for fraud risk assessment of the client.
2. The fraud risk evaluation module adopts a fraud risk analysis module based on big data support to carry out fraud risk analysis on the collected customer information, so that the fraud risk evaluation report obtained through analysis is more accurate. The customer information that presents a fraud risk facilitates the anti-fraud prediction module to calculate a fraud risk score for the customer as a whole.
3. The fraud risk score calculated by the fraud risk weight proportion rule can be used for measuring the fraud of the customer integrally, so that the fraud risk of the customer can be evaluated visually, the bad loan is analyzed and identified, and the loss of the bad loan to a bank is reduced.
Further, the system also comprises an information marking module and a model building module; wherein:
the information marking module is used for marking the client information with the fraud label information according to a preset fraud behavior judgment rule;
and the model establishing module is used for carrying out model training according to the customer information and the fraud label information to obtain a fraud risk analysis model.
Has the advantages that: the traditional fraud identification is manual tagging, i.e. a human being looks into the fraud client to determine whether the client is a fraudulent client, which is more accurate but at a considerable cost in terms of manpower and time. The system judges and marks the client information with the fraudulent behavior by using the fraudulent behavior judgment rule, and then estimates whether the client has the anti-fraud risk or not in the client trust phase by combining the machine learning algorithm, thereby automatically identifying the client fraud risk and further reducing the loss of the bank.
The model building module is used for calculating the information missing rate according to the missing data in the client information, performing data binning on the client information if the calculated missing rate is smaller than a missing rate threshold value, discretizing the client information, obtaining the prediction capability of each piece of information in the client information by calculating the information value of each piece of information of the client information, and sending the client information of which the information value reaches a preset information value threshold value to the model building module.
Has the advantages that: 1. the method comprises the steps of conducting preliminary examination on the information perfection degree of customer information, collecting customer information which is not enough to comprehensively examine the fraud risk of customers for customers with the information perfection degree lower than a set information loss rate threshold, directly skipping an anti-fraud prediction module to conduct fraud risk prediction without applying loan, saving an anti-fraud risk examination period on one hand, and avoiding fraud risk evaluation errors of a fraud risk analysis model due to insufficient customer information on the other hand.
2. And for the client information with the deletion rate lower than the deletion rate threshold, performing data binning on the client information to discretize the client information, so that the rapid iteration of a fraud risk analysis model is facilitated, and the fraud risk assessment efficiency is improved.
3. And selecting the size of the prediction capability of the information value measurement information in the customer information, and screening out the customer information larger than the threshold value of the information value, so that the fraud risk evaluation result of the later stage fraud risk evaluation module through the fraud risk analysis model is more accurate.
Further, still include customer information receiving module and postpone repayment and examine module, wherein:
the client information receiving module is used for receiving a delayed repayment application request sent by a client terminal, the delayed repayment application request comprises identity information, a repayment amount, a delayed repayment reason and a delayed repayment deadline, the identity information comprises a job company, and the delayed repayment reason is payroll delay;
the client information acquisition module is also used for acquiring the repayment condition information of other clients of the same employment company from a pre-stored client information base according to the employment company;
and the delayed repayment auditing module is used for calculating a repayment rate according to the obtained repayment condition information, and if the repayment rate is lower than a repayment rate threshold value, auditing the delayed repayment request.
Has the advantages that: the method is characterized in that whether a delayed repayment application request submitted by a client is real or not is judged by combining repayment conditions of other employees of the same employment company, some clients can still continue repayment when the clients have surplus in hands during the payroll delay, but most people still use the same month of wages for repayment, the loan often cannot be repayed on time when the payroll delay is carried out, so that the repayment rate of the timely repayment of the employees of the same employment company is calculated, the company is considered to delay the payroll when the repayment rate is lower than a repayment rate threshold value, and the influence of the non-subjective factor on personal credit of the client is avoided by delaying the delayed repayment application request, and expected fraud risk behaviors are generated for the clients.
Further, the customer information also includes a stock relation company;
the client information acquisition module is also used for acquiring repayment condition information of the related working clients of the equity relationship company according to the client information.
Has the advantages that: the income of some clients not only lies in the wages but also lies in the dividend of equity, and whether the clients can not pay on time due to the extension of the wages can be judged more comprehensively by judging whether the employees in the company related to the equity relation of the clients delay the wages or not.
Further, the client information acquisition module is also used for capturing the consumption amount of the client corresponding to the latest on-time repayment period and the current repayment period and the average wage within a preset time period according to the identity information;
and the delayed repayment auditing module is also used for comparing and analyzing the on-time repayment period with the consumption amount corresponding to the current repayment period, and if the consumption amount corresponding to the current repayment period is larger than the consumption amount of the on-time repayment period and the consumption amount difference value between the two is lower than the dynamic amount threshold value, the dynamic amount threshold value is dynamically set according to the average wage and the repayment amount in the preset time period and the preset dynamic amount threshold value setting rule through the delayed repayment request.
Has the advantages that: some customers cannot pay on time due to too large consumption in the month, subjective overdue payment behaviors exist in the behavior of the customers, and statistics of the overdue payment behaviors can be avoided when wages cannot be issued on time right, so that fraud risks are generated. According to the technical scheme, whether excess consumption exists in the client in the current repayment period is judged by comparing and analyzing the current consumption amount of the client with the consumption amount in the on-time repayment period, and the fluctuating consumption amount (dynamic amount threshold value) of consumption is set through the average wage and the repayment amount of the client so as to adapt to consumption differences of different income groups, so that the excess consumption condition is judged more accurately.
The multi-rule anti-fraud prediction method based on the user behavior data comprises the following steps:
the information acquisition step, namely acquiring customer information through a customer information acquisition module, wherein the customer information comprises historical transaction behaviors, personal credit, personal condition information, identity information and three-party data;
a fraud risk evaluation step, wherein the obtained customer information is subjected to fraud risk analysis through a fraud risk analysis model to obtain a fraud risk evaluation report of the customer, and the fraud risk evaluation report comprises the customer information of the existing fraud risk and corresponding fraud risk analysis information;
and in the anti-fraud prediction step, calculating the fraud risk score of the whole customer according to the preset fraud risk weight proportion rule and the customer information of the fraud risk in the fraud risk evaluation report, wherein the fraud risk weight proportion rule is a weight proportion set according to the influence importance degree of each piece of information in the customer information on the fraud risk.
Has the advantages that: by the method, fraud risk assessment is carried out on the multiple dimensions of historical transaction behaviors, personal credit, personal condition information, identity information and the like of the client according to a fraud risk analysis model generated under the support of fraud risk big data, so that risk assessment can be carried out more accurately. And the fraud risk assessment report is quantified through the fraud risk weight proportion rule, so that the fraud risk degree of a client can be conveniently and intuitively known, and whether the loan application is passed or not is judged.
Further, an information marking step and a model establishing step are also included between the information acquisition step and the fraud risk assessment step,
wherein:
marking the information, namely marking the client information with fraud label information according to a preset fraud behavior judgment rule;
and a model establishing step, namely performing model training according to the customer information and the fraud label information to obtain a fraud risk analysis model.
Has the advantages that: by the method, the data sample for establishing the model is more accurate, and a fraud risk assessment report obtained by analyzing the fraud risk analysis model is more accurate.
Further, the method executed in parallel with the information marking step and the model building step further comprises a data screening step, wherein the data screening step specifically comprises the following steps:
step S1, calculating the information missing rate according to the missing data in the customer information, and if the calculated missing rate is less than the missing rate threshold, executing step S2;
and step S2, performing data binning on the client information, discretizing the client information, calculating the information value of each piece of the client information to obtain the prediction capability of each piece of the client information, and sending the client information of which the information value reaches a preset information value threshold value to the model building module.
Has the advantages that: by the method, the loan qualification of the client can be preliminarily screened, and the fraud risk judgment period is shortened. And the client information which preliminarily accords with the loan qualification is preprocessed, so that the establishment of the fraud risk which is beneficial to a fraud risk analysis model is screened out.
Further, after the anti-fraud prediction step, the method further comprises the following steps:
a request receiving step, namely receiving a delayed repayment application request sent by a client terminal, wherein the delayed repayment application request comprises identity information, a repayment amount, a delayed repayment reason and a delayed repayment deadline, the identity information comprises a employment company, and the delayed repayment reason is payroll delay;
the information acquisition step, namely acquiring the repayment condition information of other clients of the same working company from a pre-stored client information base according to the working company and acquiring the repayment condition information of the related working client of the equity relationship company; capturing the consumption amount of the customer corresponding to the latest on-time repayment period and the current repayment period and the average wage in a preset time period according to the identity information;
a dynamic amount threshold setting step, wherein a dynamic amount threshold is dynamically set according to the average wage and repayment amount in a preset time period and a preset dynamic amount threshold setting rule;
and a delayed repayment auditing step, namely calculating a repayment rate according to the obtained repayment situation information, comparing and analyzing the consumption amount corresponding to the on-time repayment period and the current repayment period if the repayment rate is lower than a repayment rate threshold, and auditing the delayed repayment request if the consumption amount corresponding to the current repayment period is larger than the consumption amount of the on-time repayment period and the consumption amount difference between the two is lower than a dynamic amount threshold.
Has the advantages that: by the method, the fact that whether the client is the reason why the client can not pay on time due to the fact that the client objectively pays due to the delayed salary can be analyzed from the staff repayment condition of the same working company of the client and the current consumption condition of the client, so that the client can not be involved in overdue repayment and the personal credit problem caused by the overdue repayment due to the objective reason of the delayed salary, and the malicious behavior that the client cheats to pay off due to the overdue repayment due to the high-volume consumption due to the overdue behavior of the overdue repayment is avoided.
Drawings
FIG. 1 is a logic diagram of a first embodiment of a multi-rule anti-fraud prediction method and system based on user behavior data;
FIG. 2 is a flowchart of a first embodiment of a method and system for multi-rule anti-fraud prediction based on user behavior data;
FIG. 3 is a logic diagram of a second embodiment of a multi-rule anti-fraud prediction method and system based on user behavior data;
FIG. 4 is a flowchart of a second embodiment of a multi-rule anti-fraud prediction method and system based on user behavior data.
Detailed Description
The following is further detailed by way of specific embodiments:
example one
The multi-rule anti-fraud prediction system based on user behavior data, as shown in fig. 1, includes a customer information acquisition module, an information marking module, a model building module, a feature engineering module, a fraud risk assessment module and an anti-fraud prediction module, wherein:
the client information acquisition module is used for acquiring client information, and the client information comprises historical transaction behaviors, personal credit investigation, personal condition information, identity information and three-party data;
the information marking module is used for marking the client information with the fraud label information according to a preset fraud behavior judgment rule; the fraud tag information Y is multi-regular. The traditional fraud judgment rule is to only use the fact that the number of overdue days (current/historical) of the client is larger than a fixed threshold, but neglect the fraud condition that the client borrows a plurality of times in a short time and is short or the past fraud times are short. Therefore, the fraud judgment rule in the technical scheme fully considers the multi-dimensional fraud condition of the client, and specifically comprises the following steps: the fraud judgment rule is as follows: the maximum historical expected (including) days of the customer exceeds 90 days, and the current expected days of the first borrowing exceeds 30 days; the first borrowing has overdue behavior, and the borrowing times exceed 4 within 15 days after the first borrowing; the customer first borrows the cumulative expected number of days for the first three periods to exceed 30 days.
And the model establishing module is used for carrying out model training according to the customer information and the fraud label information to obtain a fraud risk analysis model. Specifically, the training model may adopt a lasso-logistic model and may also adopt an XGBoost model, the lasso-logistic model is used for predicting the probability that the customer is a fraudulent customer, and the XGBoost model is used for predicting whether the customer is a fraudulent customer, and the lasso-logistic model is preferred in this embodiment.
The characteristic engineering module is used for calculating an information loss rate according to the loss data in the client information, carrying out data binning on the client information if the calculated loss rate is smaller than a loss rate threshold value, discretizing the client information, calculating the information value of each piece of the client information to obtain the prediction capability of each piece of the client information, and sending the client information of which the information value reaches a preset information value threshold value to the model establishing module. In this embodiment, the loss rate threshold is 90%, and the information value threshold (IV) is 0.1. Because the basic data of the three-party data source is more, irregular and complex, how to construct and select the fraud variable is a solution point of the technical scheme, the specific process of data screening in the embodiment is as follows: 1. segmenting variables in the client information according to rows/columns, wherein the continuous variables in the client information calculate the mean value and the median, and the discrete variables are counted; 2. the method comprises the steps of carrying out long-widening on variables in client information, specifically carrying out dictionary type variables, using regular matching, extracting discrete counting variables, carrying out variable binning and regrouping on discrete calculation variables, finally generating more than 600 derivative variables, connecting the obtained derivative variables with fraud tag information, calculating the IV values of the obtained derivative variables, screening out 30 pieces of client information larger than the information value threshold value, and sending the client information to a model building module for self-learning of a risk analysis model.
The fraud risk evaluation module is used for carrying out fraud risk analysis on the obtained customer information through a fraud risk analysis model to obtain a fraud risk evaluation report of the customer, wherein the fraud risk evaluation report comprises the customer information of the existing fraud risk and corresponding fraud risk analysis information.
And the fraud prevention prediction module is used for calculating the fraud risk score of the whole customer according to the preset fraud risk weight proportion rule and the customer information of the fraud risk in the fraud risk evaluation report, wherein the fraud risk weight proportion rule is a weight proportion set according to the influence importance degree of each information in the customer information on the fraud risk.
As shown in fig. 2, the method applied to the above system includes the following steps:
the information acquisition step, namely acquiring customer information through a customer information acquisition module, wherein the customer information comprises historical transaction behaviors, personal credit, personal condition information, identity information and three-party data;
marking the information, namely marking the client information with fraud label information according to a preset fraud behavior judgment rule;
and a model establishing step, namely performing model training according to the customer information and the fraud label information to obtain a fraud risk analysis model.
A fraud risk evaluation step, wherein the obtained customer information is subjected to fraud risk analysis through a fraud risk analysis model to obtain a fraud risk evaluation report of the customer, and the fraud risk evaluation report comprises the customer information of the existing fraud risk and corresponding fraud risk analysis information;
and in the anti-fraud prediction step, calculating the fraud risk score of the whole customer according to the preset fraud risk weight proportion rule and the customer information of the fraud risk in the fraud risk evaluation report, wherein the fraud risk weight proportion rule is a weight proportion set according to the influence importance degree of each piece of information in the customer information on the fraud risk.
In addition, the method executed in parallel with the information marking step and the model building step further comprises a data screening step, and the data screening step specifically comprises the following steps:
step S1, calculating the information missing rate according to the missing data in the customer information, and if the calculated missing rate is less than the missing rate threshold, executing step S2;
and step S2, performing data binning on the client information, discretizing the client information, calculating the information value of each piece of the client information to obtain the prediction capability of each piece of the client information, and sending the client information of which the information value reaches a preset information value threshold value to the model building module.
Example two
The difference from the first embodiment is that: as shown in fig. 3, the system further comprises a client information receiving module and a delayed repayment auditing module, wherein:
the client information receiving module is used for receiving a delayed repayment application request sent by a client terminal, the delayed repayment application request comprises identity information, a repayment amount, a delayed repayment reason and a delayed repayment deadline, the identity information comprises a employment company and a stock right relationship company, and the delayed repayment reason is payroll delay;
the client information acquisition module is also used for acquiring the repayment condition information of other clients of the same employment company from a pre-stored client information base according to the employment company; the client information acquisition module is also used for acquiring repayment condition information of the related working clients of the equity relationship company according to the client information. The client information acquisition module is also used for capturing the consumption amount of the client corresponding to the latest on-time repayment period and the current repayment period and the average wage in a preset time period according to the identity information;
and the delayed repayment auditing module is used for calculating a repayment rate according to the obtained repayment condition information, and if the repayment rate is lower than a repayment rate threshold value, auditing the delayed repayment request. And the delayed repayment auditing module is also used for comparing and analyzing the on-time repayment period with the consumption amount corresponding to the current repayment period, and if the consumption amount corresponding to the current repayment period is larger than the consumption amount of the on-time repayment period and the consumption amount difference value between the two is lower than the dynamic amount threshold value, the dynamic amount threshold value is dynamically set according to the average wage and the repayment amount in the preset time period and the preset dynamic amount threshold value setting rule through the delayed repayment request.
The method applied to the system, after the anti-fraud prediction step, as shown in fig. 4, further comprises the steps of:
a request receiving step, namely receiving a delayed repayment application request sent by a client terminal, wherein the delayed repayment application request comprises identity information, a repayment amount, a delayed repayment reason and a delayed repayment deadline, the identity information comprises a employment company, and the delayed repayment reason is payroll delay;
the information acquisition step, namely acquiring the repayment condition information of other clients of the same working company from a pre-stored client information base according to the working company and acquiring the repayment condition information of the related working client of the equity relationship company; capturing the consumption amount of the customer corresponding to the latest on-time repayment period and the current repayment period and the average wage in a preset time period according to the identity information;
a dynamic amount threshold setting step, wherein a dynamic amount threshold is dynamically set according to the average wage and repayment amount in a preset time period and a preset dynamic amount threshold setting rule;
and a delayed repayment auditing step, namely calculating a repayment rate according to the obtained repayment situation information, comparing and analyzing the consumption amount corresponding to the on-time repayment period and the current repayment period if the repayment rate is lower than a repayment rate threshold, and auditing the delayed repayment request if the consumption amount corresponding to the current repayment period is larger than the consumption amount of the on-time repayment period and the consumption amount difference between the two is lower than a dynamic amount threshold.
The foregoing is merely an example of the present invention and common general knowledge of known specific structures and features of the embodiments is not described herein in any greater detail. It should be noted that, for those skilled in the art, without departing from the structure of the present invention, several changes and modifications can be made, which should also be regarded as the protection scope of the present invention, and these will not affect the effect of the implementation of the present invention and the practicability of the patent. The scope of the claims of the present application shall be determined by the contents of the claims, and the description of the embodiments and the like in the specification shall be used to explain the contents of the claims.
Claims (10)
1. The multi-rule anti-fraud prediction system based on user behavior data is characterized by comprising a client information acquisition module, a fraud risk assessment module and an anti-fraud prediction module, wherein:
the client information acquisition module is used for acquiring client information, wherein the client information comprises historical transaction behaviors, personal credit investigation, personal condition information, identity information and three-party data;
the fraud risk evaluation module is used for carrying out fraud risk analysis on the obtained client information through a fraud risk analysis model to obtain a fraud risk evaluation report of the client, wherein the fraud risk evaluation report comprises the client information of the existing fraud risk and corresponding fraud risk analysis information;
and the fraud prevention prediction module is used for calculating the fraud risk score of the whole customer according to the preset fraud risk weight proportion rule and the customer information of the fraud risk in the fraud risk evaluation report, wherein the fraud risk weight proportion rule is a weight proportion set according to the influence importance degree of each information in the customer information on the fraud risk.
2. The multi-rule anti-fraud prediction system based on user behavior data of claim 1, characterized by: the system also comprises an information marking module and a model building module; wherein:
the information marking module is used for marking the client information with the fraud label information according to a preset fraud behavior judgment rule;
and the model establishing module is used for carrying out model training according to the customer information and the fraud label information to obtain a fraud risk analysis model.
3. The multi-rule anti-fraud prediction system based on user behavior data of claim 1, characterized by: the fraud prevention prediction system further comprises a characteristic engineering module, wherein the characteristic engineering module is used for calculating an information loss rate according to the loss data in the customer information, performing data binning on the customer information if the calculated loss rate is smaller than a loss rate threshold value, discretizing the customer information, obtaining the prediction capability of each piece of information in the customer information by calculating the information value of each piece of information of the customer information, and sending the customer information of which the information value reaches a preset information value threshold value to the fraud prevention prediction module.
4. The multi-rule anti-fraud prediction system based on user behavior data of claim 1, characterized by: still include customer information receiving module and postpone repayment and examine module, wherein: the client information receiving module is used for receiving a delayed repayment application request sent by a client terminal, the delayed repayment application request comprises identity information, a repayment amount, a delayed repayment reason and a delayed repayment deadline, the identity information comprises a job company, and the delayed repayment reason is payroll delay;
the client information acquisition module is also used for acquiring the repayment condition information of other clients of the same employment company from a pre-stored client information base according to the employment company;
and the delayed repayment auditing module is used for calculating a repayment rate according to the obtained repayment condition information, and if the repayment rate is lower than a repayment rate threshold value, auditing the delayed repayment request.
5. The multi-rule anti-fraud prediction system based on user behavior data of claim 4, characterized in that: the customer information also includes a equity relationship company;
the client information acquisition module is also used for acquiring repayment condition information of the related working clients of the equity relationship company according to the client information.
6. The multi-rule anti-fraud prediction system based on user behavior data of claim 4, characterized in that: the client information acquisition module is also used for capturing the consumption amount of the client corresponding to the latest on-time repayment period and the current repayment period and the average wage in a preset time period according to the identity information;
and the delayed repayment auditing module is also used for comparing and analyzing the on-time repayment period with the consumption amount corresponding to the current repayment period, and if the consumption amount corresponding to the current repayment period is larger than the consumption amount of the on-time repayment period and the consumption amount difference value between the two is lower than the dynamic amount threshold value, the dynamic amount threshold value is dynamically set according to the average wage and the repayment amount in the preset time period and the preset dynamic amount threshold value setting rule through the delayed repayment request.
7. The multi-rule anti-fraud prediction method based on the user behavior data is characterized by comprising the following steps of:
the information acquisition step, namely acquiring customer information through a customer information acquisition module, wherein the customer information comprises historical transaction behaviors, personal credit, personal condition information, identity information and three-party data;
a fraud risk evaluation step, wherein the obtained customer information is subjected to fraud risk analysis through a fraud risk analysis model to obtain a fraud risk evaluation report of the customer, and the fraud risk evaluation report comprises the customer information of the existing fraud risk and corresponding fraud risk analysis information;
and in the anti-fraud prediction step, calculating the fraud risk score of the whole customer according to the preset fraud risk weight proportion rule and the customer information of the fraud risk in the fraud risk evaluation report, wherein the fraud risk weight proportion rule is a weight proportion set according to the influence importance degree of each piece of information in the customer information on the fraud risk.
8. The multi-rule anti-fraud prediction method based on user behavior data according to claim 7, characterized in that: an information marking step and a model establishing step are further included between the information obtaining step and the fraud risk assessment step, wherein:
marking the information, namely marking the client information with fraud label information according to a preset fraud behavior judgment rule;
and a model establishing step, namely performing model training according to the customer information and the fraud label information to obtain a fraud risk analysis model.
9. The multi-rule anti-fraud prediction method based on user behavior data according to claim 7, characterized in that: the method, executed in parallel with the information marking step and the model building step, further comprises a data screening step, wherein the data screening step specifically comprises the following steps:
step S1, calculating the information missing rate according to the missing data in the customer information, and if the calculated missing rate is less than the missing rate threshold, executing step S2;
and step S2, performing data binning on the client information, discretizing the client information, calculating the information value of each piece of the client information to obtain the prediction capability of each piece of the client information, and sending the client information of which the information value reaches a preset information value threshold value to the model building module.
10. The multi-rule anti-fraud prediction method based on user behavior data according to claim 7, characterized in that: after the anti-fraud prediction step, the method further comprises the following steps:
a request receiving step, namely receiving a delayed repayment application request sent by a client terminal, wherein the delayed repayment application request comprises identity information, a repayment amount, a delayed repayment reason and a delayed repayment deadline, the identity information comprises a employment company, and the delayed repayment reason is payroll delay;
the information acquisition step, namely acquiring the repayment condition information of other clients of the same working company from a pre-stored client information base according to the working company and acquiring the repayment condition information of the related working client of the equity relationship company; capturing the consumption amount of the customer corresponding to the latest on-time repayment period and the current repayment period and the average wage in a preset time period according to the identity information;
a dynamic amount threshold setting step, wherein a dynamic amount threshold is dynamically set according to the average wage and repayment amount in a preset time period and a preset dynamic amount threshold setting rule;
and a delayed repayment auditing step, namely calculating a repayment rate according to the obtained repayment situation information, comparing and analyzing the consumption amount corresponding to the on-time repayment period and the current repayment period if the repayment rate is lower than a repayment rate threshold, and auditing the delayed repayment request if the consumption amount corresponding to the current repayment period is larger than the consumption amount of the on-time repayment period and the consumption amount difference between the two is lower than a dynamic amount threshold.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011631672.4A CN112700321A (en) | 2020-12-31 | 2020-12-31 | Multi-rule anti-fraud prediction method and system based on user behavior data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011631672.4A CN112700321A (en) | 2020-12-31 | 2020-12-31 | Multi-rule anti-fraud prediction method and system based on user behavior data |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112700321A true CN112700321A (en) | 2021-04-23 |
Family
ID=75513555
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011631672.4A Pending CN112700321A (en) | 2020-12-31 | 2020-12-31 | Multi-rule anti-fraud prediction method and system based on user behavior data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112700321A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112991052A (en) * | 2021-04-25 | 2021-06-18 | 大箴(杭州)科技有限公司 | Repayment capability evaluation method and device |
| CN113392422A (en) * | 2021-08-16 | 2021-09-14 | 华控清交信息科技(北京)有限公司 | Data processing method and device and data processing device |
| CN113610534A (en) * | 2021-07-28 | 2021-11-05 | 浙江惠瀜网络科技有限公司 | Data processing method and device for anti-fraud |
| CN113837866A (en) * | 2021-09-29 | 2021-12-24 | 重庆富民银行股份有限公司 | Two-stage management method and system based on full stock customer |
| CN113987182A (en) * | 2021-10-28 | 2022-01-28 | 深圳永安在线科技有限公司 | Fraud entity identification method, device and related equipment based on security intelligence |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104408610A (en) * | 2014-12-03 | 2015-03-11 | 苏州贝多环保技术有限公司 | Third-party payment platform business processing method based on risk assessment |
| CN108053310A (en) * | 2017-11-24 | 2018-05-18 | 深圳市牛鼎丰科技有限公司 | Credit scoring method, apparatus, computer equipment and storage medium |
| CN109034502A (en) * | 2018-09-04 | 2018-12-18 | 中国光大银行股份有限公司信用卡中心 | Anti- Fraud Prediction method and device |
| CN109035003A (en) * | 2018-07-04 | 2018-12-18 | 北京玖富普惠信息技术有限公司 | Anti- fraud model modelling approach and anti-fraud monitoring method based on machine learning |
| CN109360084A (en) * | 2018-09-27 | 2019-02-19 | 平安科技（深圳）有限公司 | Appraisal procedure and device, storage medium, the computer equipment of reference default risk |
| CN110223165A (en) * | 2019-06-14 | 2019-09-10 | 哈尔滨哈银消费金融有限责任公司 | A kind of anti-fraud and credit risk forecast method and system based on related network |
| CN110245875A (en) * | 2019-06-21 | 2019-09-17 | 深圳前海微众银行股份有限公司 | Risk of fraud appraisal procedure, device, equipment and storage medium |
| CN111222976A (en) * | 2019-12-16 | 2020-06-02 | 北京淇瑀信息科技有限公司 | Risk prediction method and device based on network diagram data of two parties and electronic equipment |
-
2020
- 2020-12-31 CN CN202011631672.4A patent/CN112700321A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104408610A (en) * | 2014-12-03 | 2015-03-11 | 苏州贝多环保技术有限公司 | Third-party payment platform business processing method based on risk assessment |
| CN108053310A (en) * | 2017-11-24 | 2018-05-18 | 深圳市牛鼎丰科技有限公司 | Credit scoring method, apparatus, computer equipment and storage medium |
| CN109035003A (en) * | 2018-07-04 | 2018-12-18 | 北京玖富普惠信息技术有限公司 | Anti- fraud model modelling approach and anti-fraud monitoring method based on machine learning |
| CN109034502A (en) * | 2018-09-04 | 2018-12-18 | 中国光大银行股份有限公司信用卡中心 | Anti- Fraud Prediction method and device |
| CN109360084A (en) * | 2018-09-27 | 2019-02-19 | 平安科技（深圳）有限公司 | Appraisal procedure and device, storage medium, the computer equipment of reference default risk |
| CN110223165A (en) * | 2019-06-14 | 2019-09-10 | 哈尔滨哈银消费金融有限责任公司 | A kind of anti-fraud and credit risk forecast method and system based on related network |
| CN110245875A (en) * | 2019-06-21 | 2019-09-17 | 深圳前海微众银行股份有限公司 | Risk of fraud appraisal procedure, device, equipment and storage medium |
| CN111222976A (en) * | 2019-12-16 | 2020-06-02 | 北京淇瑀信息科技有限公司 | Risk prediction method and device based on network diagram data of two parties and electronic equipment |
Non-Patent Citations (1)
| Title |
|---|
| 王艳辉: "《文本挖掘中若干关键问题研究》", 中国科学技术大学出版社, pages: 34 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112991052A (en) * | 2021-04-25 | 2021-06-18 | 大箴(杭州)科技有限公司 | Repayment capability evaluation method and device |
| CN112991052B (en) * | 2021-04-25 | 2022-01-25 | 大箴(杭州)科技有限公司 | Repayment capability evaluation method and device |
| CN113610534A (en) * | 2021-07-28 | 2021-11-05 | 浙江惠瀜网络科技有限公司 | Data processing method and device for anti-fraud |
| CN113392422A (en) * | 2021-08-16 | 2021-09-14 | 华控清交信息科技(北京)有限公司 | Data processing method and device and data processing device |
| CN113837866A (en) * | 2021-09-29 | 2021-12-24 | 重庆富民银行股份有限公司 | Two-stage management method and system based on full stock customer |
| CN113987182A (en) * | 2021-10-28 | 2022-01-28 | 深圳永安在线科技有限公司 | Fraud entity identification method, device and related equipment based on security intelligence |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Gande et al. | CEO compensation and risk-taking at financial firms: Evidence from US federal loan assistance | |
| CN112700321A (en) | Multi-rule anti-fraud prediction method and system based on user behavior data | |
| Raghavan | Risk management in banks | |
| Ongena et al. | Creditor concentration: An empirical investigation | |
| Han et al. | Effects of debt collection practices on loss given default | |
| Kim et al. | Financial statement comparability and managers’ use of corporate resources | |
| Ayako et al. | Determinants of firm value in Kenya: Case of commercial banks listed at the Nairobi Securities Exchange | |
| Al-Saleh et al. | Prediction of financial distress for commercial banks in Kuwait | |
| Wang et al. | Can gold hedge against oil price movements: Evidence from GARCH-EVT wavelet modeling | |
| CN111861729A (en) | Behavior scoring system and method based on lstm | |
| Ufo et al. | Determinants of financial distress in manufacturing firms of Ethiopia | |
| Šverko Grdić et al. | Insolvency in the Republic of Croatia | |
| Samreen et al. | Design and Development of Credit Scoring Model for the Commercial Banks in Pakistan: Forecasting Creditworthiness of Corporate Borrowers | |
| Chen et al. | Modeling price and variance jump clustering using the marked hawkes process | |
| CN115564551A (en) | Enterprise credit rating method for financial big data | |
| Gupta et al. | Indian Banks and Basel-Ii-An Econometric Analysis | |
| Comporek | The use of operational cash flow in the estimation of accrual-based earnings management | |
| Türkcan | Financial Failure Prediction in Banks: The Case of European Union Countries | |
| Daoud et al. | The Econometrics Effect of Information Technology Investment on Financial Performance in the Jordanian Banking Sector over the Period 1993-2014 | |
| Breeden et al. | Impacts of Covid-19 on model risk management | |
| Khumalo et al. | Evaluating the credit risk and macroeconomic interaction in South African Banks | |
| Ufo | Impact of Financial Distress on the Liquidity of Selected Manufacturing Firms of Ethiopia | |
| Cruz Rambaud et al. | Risk-based price of mortgages: an assessment of the loss due to the withdrawal of floor clauses in Spain | |
| Drábková | Fraud risk management from the perspective of CFEBT risk triangle of accounting errors and frauds | |
| Sepehrdoust et al. | Credit risk management of commercial banks in Iran: using logistic model |
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 | ||
| CB03 | Change of inventor or designer information | ||
| CB03 | Change of inventor or designer information |
Inventor after: Yang Fang Inventor after: Xu Xiaoming Inventor after: Ying Jie Inventor after: Chen Chijie Inventor before: Yang Fang Inventor before: Xu Xiaoming Inventor before: Ying Jie Inventor before: Chen Chi |
|
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210423 |