Methods for identifying customer quality in automated credit systems using visual scoring

With the development of microfinance in the Russian Federation, the direction of credit scoring is gaining more and more urgency. In the current conditions of close competition and high regulatory burden on the part of the Central Bank, microfinance organizations are beginning to apply increasingly unconventional tools for enriching data on the borrower. Such a tool became visual scoring. In this article, the author gives examples of the practices of visual scoring used by online MFIs and shares the results of his research, according to which the hypothesis about the possibility of a connection between the features of the borrower's face and the probability of overdue debt on the loan is allowed. All analysis and conclusions to the article are the result of the application of machine learning technologies, in-depth training and basic algorithms of computer vision available to the average user.