As superintendency development, the net borrows an industry to be lifted once more shuffle upsurge. Failure of transition of platform of the conduct financial transactions below the line, small platform can'ts bear weigh exiting the market to compete in succession, the net borrowed industry competition to enter realer phase.
"824" net borrows superintendency plan to put forward, "Direct or covert offer to lender assure or acceptance protects this protecting breath " ; "57 article " puts forward clearly, "Prohibit the orgnaization inside administer continues to add a risk extraction, newly to pay gold fully. "These are meant, superintendency principle is very clear, the user should have risk consciousness, platform won't help your disclose the whole inside story. But user oneself hard control a venture, even if can see the information of borrower, also cannot reach objective and true assessment. Below this setting, the wind of platform accuses actual strength to become those who judge a risk is only standard approximately, also be the core element of user choice platform.
Meanwhile, the user that the net borrows an orgnaization to serve often is the group that the bank is done not have or cannot cover. Data shows, in the Central Bank the person that the report that sign a letter can check inside the record that sign a letter occupies 1.4 billion population of Chinese only 25% the left and right sides. That is opposite in American the corresponding period the number of mark is 60-70% left and right sides probably. Differred between this scale probably 100 million population of 5-6. Since the bank cannot give a letter, the net borrows an orgnaization to need to build effective wind to accuse a model by oneself.
Why to accuse a model for effective wind? How does ability eliminate have potential con group, the health that assures platform is dovish. We are with foreign money pot exemple, those things that wind charges say for.
Choose foreign money pot, it is OK to basically depend on this platform accomplish 3 minutes finish examine and verify and do not need group of giant examine and verify. In 3 years of many time, accumulative total trades the forehead had been achieved nearly 55 billion, lend user exceeds 1.3 million, pen of accumulative total loan counts 26.46 million, platform maintains dovish operation. This makes clear, platform has a powerful wind to accuse a system to do prop up, and had gotten sufficient confirm, its practice is having important draw lessons froming action.
So, where is the recipe that wind of foreign money pot accuses? We are main from ask for a letter to evaluate, serve false instead and 3 respects discuss monitoring early-warning.
Big data and borrower tall dimension spend weak variable analysis to build a model
Before leasehold happening, platform should undertake asking for believing assessment to borrower above all. Traditional banking orgnaization often is to inquire background of record of salary running water, social security, job, family is waited a moment, but on one hand, the net lends the group that platform covers, these information are not complete clarity to see. On the other hand, even if user provides these information, also can't draw the outline of a clear borrower picture completely, the cost that makes a holiday is not high also. In mobile Internet times, data dimension is spent and processing computation model becomes all in all.
The practice of foreign money pot is, besides the Internet reptile technology that passes through in industry and technology of online identity attestation, just still build cooperation with tens of domestic data collaboration on the line, through pure on-line means can take the variable that takes thousands of dimension to spend. Geng Bo of CTO of pot of the cash that occupy ocean introduces, the information that relies on an user to go up in mobile Internet, the content that foreign money pot can change weak relation, fragment forms a Gao Wei to spend, fine grained changes depict. Be based on this to depict, machine study algorithm can is opposite further the user undertakes classified.
That is to say, the big data wind of foreign money pot accuses to see " big " not merely, the key is to see a machine whether process these data well. Foreign money pot relies on huge data, in the light of sincere letter, break couplet, correlation to get together kind, the many dimension such as credit spent compose to build a certain number of machine study model, can of accurate differentiate user mix false credit risk grade. Occupy Geng Bo to divulge, machine study algorithm can form the classification that a fine grained spends to borrower, the weak variable that spends through Gao Wei undertakes the model is analysed building, do not satisfy the borrower of the requirement to credit, platform can seasonable eliminate.
The credence of borrower is probability sex, the degree of confidence that reaction borrower issues in certain probability. Information dimension spends variable of wider, data more, machine study is more accurate, the granularity that depicts to borrower is finer, the cost that makes a holiday is higher, the degree of confidence that gives a letter to borrower is taller. This foundation job has been done, turn over con job to become success will come when conditions are ripe.
Con and individual discriminate with the trends of the group
Below normal circumstance, the account that bad perhaps Zhang produces leasehold exceed the time limit, besides borrower neglectful, no more than is mixed without reimbursement ability namely without reimbursement apiration. Be aimed at issue of this two sides, the wind of foreign money pot accuses basic principle is, make sure borrower has reimbursement capacity, hold off without reimbursement apiration borrower.
On reimbursement ability, the asset of foreign money pot basically comes from small consume kind of asset, small credit is commonly hundreds of, thousands of small loan, this can see from platform data. The operation data of foreign money pot shows, specified amount of loan of accumulative total of platform average per capita has 5000 only multivariate (delimit key: Accumulative total) , before 10 big borrower are waited for return amount to occupy than only 0.014% . The biggest onefold borrower is waited for return amount to occupy than only 0.001% , reflected small and dispersive characteristic well. Small and dispersive loan means can assure, if regular borrow or lend money needs, borrower can assure basically to have reimbursement capacity.
So, core problem is centered on the assessment of reimbursement apiration, detect namely actually con user. Con user cent is mixed false to form a delegation individual and con.
In individual and con respect, the phase of the data that sign a letter of data of will own research and development and other stage combines foreign money pot, build a huge database, thereby can of accurate differentiate user mix false credit risk grade. According to Introduction Geng Bo, individual and con belong to be based on Internet classically big data and have supervise / the problem that the machine study technique that supervises partly goes solving, accumulate ceaselessly as data bulk, disentombed Internet variable is progressively grow in quantity, can last iteration is optimized ceaselessly.
The difficulty with con group depends on, although it is individual and con gather, but need platform both neither is met accidentally injure is high grade borrower, can dig disinter thoroughly again all and con individual. Picture of borrower of this hang on and get together kind of algorithmic model. The practice of foreign money pot is, whether to agree with its Internet picture through information of individual of comparative database differentiate; Live body identifies a technology to examine whether oneself are operated; Wait for information through comparing pair of equipment and contact, the frequency of loan happening and concentration are spent wait for an element, much azimuth detects the behavior of borrower. Be based on likeness to spend to different user picture undertake getting together kind, union is different correlation concerns between the applicant, center eliminate to drop form a delegation the likelihood of bilk. According to Introduction Geng Bo, of group bilk be on guard the category of review of the educational inspector that do not have inspect in belonging to machine study, magnanimity study and get together kind of algorithm having crucial effect here. Through getting together kind of algorithm, foreign money pot can discriminate a con discharge in time, make timely response.
Accumulate in what oppose con technique on, foreign money pot can assure to there basically can be group sex on platform con, very rare and individual con.
Dynamic monitoring early-warning and intelligence are urged close
Of course, also should see, should be real time trends to the monitoring of the risk, ought to undertake to the user credit early-warning is mixed continuously the monitoring after borrowing, the venture of banking of credit aggravation etc that discovers an user in time, implementation is felt to risk posture of finance.
In this on one hand, foreign money pot ground to suit the real time supervisory system of oneself demand for service oneself, lash-up answers time round-the-clock assure to be in several minutes in, ensure service on the safe side. According to Introduction Geng Bo, daily metropolis is opposite foreign money pot user existence reimbursement and position of exceed the time limit are newer. These condition are updated can the parameter that efficient feedback replaces a model to the model, undertake OK and dynamicly accommodative when user crowd produces migratory fraudulent perhaps action to produce change.
And it is besides this, foreign money pot still is aimed at urge control research and development intelligence is urged accept a system. This department unites respect utmost ground to assure the privacy of user information and safety, complete record urges the whole process that control. On the other hand, intelligence is urged receive a system to carry IVR kind, realize the communication of machine and person, remind borrower reimbursement, offer its corresponding reimbursement way. As the development of AI technology, foreign money pot is resembling Gu Ge in implementation in that way intelligence customer service, undertake urging closing with the means of standardization, can make sure information is accurate communicate, prevent force to urge close.
The place on put together is narrated, from ask for a letter to evaluate, turn over con service and monitoring early-warning in light of 3 respects, foreign money pot has two very crucial essential factor: Technical drive, trends is adjusted. The gene with this and the most primitive platform is concerned. Group of its core technology comes from the whole world famous Internet enterprise, include Facebook, google, apple, microsoft, SAP, oracle, tecent. So, the framework accuses to go up in platform wind, the group is in put an end to as far as possible cannot control an element. For instance, in level of examine and verify, foreign money pot did not establish group of the examine and verify below the line with giant dimensions, offer the user with tall degree of confidence to give a letter with big data and computational model however. On the technology, the group is going after rigorous sex as far as possible, make sure trends is adjusted. Foreign money pot can assure tall trade cost of quantity, small staff and dovish operation, its recipe accuses very actual strength of the technology in Yu Feng.
The near future is wonderful article Q
