The concept of large financial data model, all data are credit data

&http://www.aliyun.com/zixun/aggregation/37954.html ">nbsp; So far, investments in large data areas have become hotter and more companies are being made. How many companies really use big data? I believe there is hardly much.

The most direct scenario for big data in American finance is the so-called credit rating system. The U.S. credit system is simple to assess: debt history, debt, credit history, and other related factors. All of these add up to form the existing scoring system in the United States.

In general, if the variables are put too much, the model will be more cumbersome to deal with. The main point is that its depth is more important than breadth. So the records of the past 20 years, and the last one, are not the same.

In addition, the focus on the user's history far more than the present, perhaps this person at first is a cock silk, recently suddenly made a fortune, perhaps his solvency will have a huge change, but such factors are not reflected in this? A lot of people don't know. How to put things in the vertical and horizontal breadth, this will appear relatively important.

What is the use of big data in finance? The same person is not the same in different applications and fields. For example, today in the company for 20 years, does not necessarily mean that he is a good employee, most likely he did not have the ability to change jobs. If you judge this person by another angle, your criteria and application variables should be completely changed. But it is a pity that no one weighs a person from this point of view.

Why is the wind control put in such an important angle at last? Like in China's Peer-to-peer company, 6 months or a year later, can be left One-fourth is a miracle, many peer-to-peer companies will die, or be bought and lost. As all the boom fades, wind control will be the most prominent position.

It is also important to consult the user directly. You can lie in one place, you can lie in two places, but if I have a large number of thousands of points, it is difficult to put tens of thousands of points in the contradiction between the situation, to disguise it, if it can be disguised as such, it is not a process of deception, So it's hard to get a person to make up a lie that doesn't get caught through big data.

Large data model concept, all data are credit data. All those key variables, if alone to know the proposed one, not much useful to judge how this person, but if all of these small factors all together, you will find that the final is very strong point, can be very accurate to determine what the person is doing. It's a very important point to look at the relationship and not see cause and effect.

The same name sounds good, machine learning, we will profoundly experience, in fact, we are very sad to urge the learning machine, is not a machine in learning us. How to communicate better with the machine, we give him a method, or give him a fact, he can be faster from the extract, more is a kind of interaction.

The second model of the big data, which we think is the source of the data. Even if the error message is information, it also embodies a person's quality.

The 3rd is the so-called modeling, in short, large data in the so-called characteristics of the change, feature extraction and the last so-called independent model details of the establishment of the final model of the integration of the previous traditional statistical theory has a great difference.

Finally this is a more interesting thing, this thing in China basically does not exist, but in the United States relatively troublesome, large data and the relationship between the relevant legislation. I believe that China's legislation in the future more and more perfect will encounter such problems, credit evaluation of some restricted zones, these prohibited areas can not touch. The first sex absolutely can not use, to decide this person in the end credit value how, this is absolutely not. Second Age, age is not as serious as sex, but the age is required, can only as a factor of addition, but not as a factor of reduction, the age of many people are not used. Third race, absolutely cannot touch the red line, can not be based on Asians, blacks, whites or Latinos, judge your credit is good or bad. For example, in China you drive in the street or in the United States, the wall is written to tell you that the place can not turn around, is because too many people in this place turned, so it will tree a brand, if this place is narrow, you do not have to write, and no one to turn around. The prohibited use of these things, in fact, the most truly embodies the essence of a person. In fact, it can be seen from our models that these forbidden things, if used, are more useful than the thousands of messages added together.

Big Data Another strange application, is that it can help you to circumvent some of the legal red line, this is not a legal ball, but because the essence of things is determined by these factors, a can highlight c,b and can highlight C,a and B must have a correlation.

Author, Gu Lingyun, Turbo Financial Group chief risk.