Shulou(Shulou.com)06/03 Report--

In the field of financial risk control, data has five attributes

1. Population attribute

Gender, age, phone number, name, home address, mainly for how to find this user

2. Consumption characteristics

What e-commerce users often buy, when to buy, and how much they consume each month can indirectly become a credit score.

For example, some customers can spend thousands of yuan a month, and the things they buy are also middle-and high-end goods. In fact, they can make a small loan.

3. Hobbies

According to the remarks on Weibo, what website you often visit, you can judge whether this person often travels and likes cars.

4. Credit attribute

Income, assets, occupation, liabilities, education, job, credit score

5. Social attributes

Through the user's moments behavior, if a person often goes in and out of some commercial and residential areas, the credit line of this person can be judged.

You can judge whether this person often changes jobs, for example, he often communicates with these people for half a year, and after another period of time, he changes jobs and exchanges with friends.

Most of the data can be divided into these five categories, the most important of which are demographic attributes and credit attributes to determine the credit line.

To determine which data is needed based on the business model, the data serves the business and risk control model

Financial risks are mainly divided into

1. Credit risk: non-repayment of borrowed money, non-repayment ability and willingness to repay

2. Fraud risk: the main challenge faced by many Internet financial companies in the early stage is fraud risk. The main reasons for fraud risk are as follows:

(1) if user data is leaked, users will impersonate other users.

(2) lack of risk control and loopholes in business. Many Internet companies have no experience in facing these risks. Banks already have a lot of experience, so many econnoisseur came to Internet finance companies to buy wool.

(3) everyone's information is asymmetric, so econnoisseur was able to cheat here and then go to another place to cheat. Econnoisseur seized the loophole in the business.

(4) Black industry chain, the emergence of financial fraud groups.

Some of the main risks in third-party payments:

1. Theft of cards and swiping, because payment institutions are all licensed institutions, and in order to protect their reputation, they will usually pay full compensation.

2. collude to defraud a loan, such as going to a beauty salon, someone will make a loan directly to the beauty salon

3. Long loans, all of which are borrowed by many platforms.

4. Account fraud

In the era of mobile Internet, people use mobile phones to complete transactions. Generally, mobile verification codes pass verification, but fraudsters pass verification by intercepting mobile phone verification codes.

It is necessary to know where the fraud occurs in order to effectively prevent the risk.

Mainly rely on big data's risk control model to land.

General risk control model:

1. Authentication model

2. Credit score model

3. Behavior scoring model

4. Fraud model

5. Third-party payment includes cash-out model and virtual transaction model (money laundering is the most likely to occur)

6. Repayment willingness model

7. Judge whether there will be excessive borrowing.

8. Identification model of vest maintenance number.

How do traditional financial institutions do risk control:

1. Traditional financial institutions will require a lot of vouchers

2. Traditional financial institutions do not have the advantage of data such as consumption characteristic data.

To judge whether it is the true will of the user himself.

Fraud risk prevention and control needs to be done at the device level. Mobile phones need to be bound, and mobile phones need to be authenticated.

Manual verification is required when logging in to the account, such as manual swiping.

Fingerprint verification, face recognition

Establish an engine to prevent fraud risk

1. Establish a database of risk characteristics: non-resident location, unusual behavior.

2. Based on abnormal behavior, the expert rule set of risk control is established.

3. Use the rule set to establish a machine learning model to make the rule set perfect continuously.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.