First, credit big data mining:
The data related to risk control in the massive big data of the Internet
E-commerce website big data: Ali, Jingdong, Suning, etc.;
Credit card website big data: I love the card, silver rate card, etc.;
Social website big data: Sina Weibo, Tencent WeChat, etc.
Big data of microfinance websites: Renren Loan, Credit Bao, etc.
Big data of payment websites: Ebay, Paypal, etc.
Big data of life service websites: PingAn Yishitong, etc. ...
Before the data processing, the understanding of the business, the understanding of the data is very important, which determines which data raw materials to be selected for data mining, in the "data factory" before the workload usually accounted for more than 60% of the entire process.
In terms of data raw materials, more and more Internet online dynamic big data is added. For example, a false borrower applicant information can be identified by analyzing online behavioral traces, a real Internet user will always leave a trace on the network. The timeliness of the data that is useful for credit is also critical, and dynamic data that is generally recognized as valid by the credit industry is usually backed up by 24 months of data from the present.
By obtaining raw big data from multiple sources and analyzing it using mathematical operations and statistical models to assess the borrower's credit risk, a typical domestic enterprise is the Shenzhourong big data risk control platform. Risk control with big data analysis is Iberia's core technology. Their raw data sources are very extensive.
The core technology and secret of their data factory is that they have developed multiple learning machine-based analytical models to analyze over 3000+ dimensions of raw information data of each credit applicant and come up with metrics that can make measurements on their behaviors, which can all be done in less than 5 seconds.
Two, risk control operations:
Pre-credit marketing: 1, existing customer development, new customer development; 2, pre-approval, application scoring 3, pre-approval, customer access, pre-credit amount estimation.
Loan approval: 1, fraud screening, anti-fraud monitoring; 2, application re-scoring; 3, credit approval; 4, loan pricing.
Post-credit management: 1, behavioral scoring model; 2, quota management; 3, risk early warning, pre-collection; 4, collection scoring, collection strategy.
At present, the online speed of loan approval has achieved a breakthrough, and the loan approval rate has been significantly improved. The same type of users, with collateral, income flow proof and other crude traditional risk control methods, the loan approval rate is about 15%, while the use of big data models combined with manual approval rate can reach more than 30%. As for the overdue rate of loans, take the 12-month default risk as an example, the overdue rate of users screened by the online credit approval model of Shenzhou Rong is half lower than that of those who have not been screened.