|
Consumer credit industry to flourish, as well as consumer credit business risks and returns that corresponds to objective laws, financial institutions have to face huge credit risk in the pursuit of huge profits at the same time, how the ship to avoid potential credit risk banks and important issue facing the credit institutions. Banks need before credit loans, credit customer a science credit assessment, objective, comprehensive and accurate assessment of the consumer's repayment ability and willingness to repay, in order to avoid, control, reduce bad debt losses. In Western countries, the personal credit score is commonly used to quantitatively evaluate consumer credit consumer credit status. The help of data mining techniques to construct credit scoring models to explore patterns and rules inherent in the data, as a basis for decision making of consumer credit management. In China, due to the imperfections of the social credit system, and the backwardness of the consumer credit industry, the development of personal credit model has only just started, the lack of experience for the development of appropriate credit model, this article will explore this. Commonly used credit scoring techniques are generally divided into statistical methods and non-statistical methods. Statistical methods, including linear regression, discriminant analysis, logistic regression, decision tree, non-statistical methods, including linear programming, neural networks, genetic algorithms, expert systems. But for these developers credit model technology, the kind of method is best, yet the same conclusion. Papers real credit data analysis, the most common use of discriminant analysis, logistic regression, neural network adaptive research. They were established model to classify customers, and compare the model performance. Comparison, various models have some predictive ability can be good or bad customer moderate areas to separate. Logistic regression models evaluated in these three technologies. Optimal model can be used by commercial banks, worthy of promotion in practice. In the process of establishing credit model, the data is the basis for the establishment of personal credit scoring model, leave the data, \The actual data collected are generally \Data pre-processing equipment includes: data cleansing, data conversion, and variable clustering, especially in-depth analysis feature items excessive categorical variables. Finally, modeling summarize, add a few comments for the actual applications. Pointed out the inadequacies of the paper, as well as in-depth study need to be in the future.
|