The Construction Of Index System Of Personal Social Credit, And Early Warning Models

These initiatives of The People's Bank of China and the government, illustrate the value of personal credit information by more and more attention and concern. However, the definition of individual credit is not clear, the index system still not complete, personal information not complete, and personal information sharing mechanism has not been formed, in addition to the early warning model selection is also insufficient.The purpose of this paper is try to design a set of relatively complete, reasonable individual credit system, including individual credit index system, quantitative scoring rules, determination of weight, combination of credit score, and the divide of credit rank. Then apply the scoring index system, and finally set up the costing early-warning model based on the ECM standard. This paper start from the personal finance credit most concerned about by banks and other lending institutions, expanding connotation of personal credit, and put forward the concept of personal social credit. Personal Social credit not only include financial credit which is concerned by banks and other lending institutions, including occupational credit concerned by employers, industry associations, and the public, but also include public credit cared about by the government and other public department. We do not confine personal credit to individual bank credit any more. Three categories of personal credit has different degree of cross, namely the different types of credit between contains the same index. But some of the intersect indicators in different credit categories are in the different weight, if the weights are not compartmentalized then it is easy to cause unreasonable scoring. Personal social credit is also closely related to individual occupation, but occupation has different influence on three classification of personal social credit. Generally the impact on personal occupation is the biggest, followed by individual public credit, the last by personal financial credit. Just because of occupation has the biggest impact on personal occupational credit and public credit, so the redesign on occupational and public credit index system of special occupation appear to particularly important. This article selects the lawyer more respective of the occupation as an amendment to personal occupation, to select the most representative occupation, such as individually-owned businessman as an amendment to personal public credit.At present, although there are some literature has been pointed out that we should use Expected cost of Misclassification (ECM) instead of Total Error Probability(TPM) as a evaluation of model predict effect, yet they convert TMP to ECM just on the base of models selected by standard of minimum TMP. I read in the literature and have not notice early warning models established directly by the criteria of the expected cost of misclassification or profit maximization or these scholars did not point out the realization way of profit scoring model.In the construction of early-warning model, this paper compares the traditional Total Probability of Misclassification (i.e., the prediction accuracy), Expected Cost of Misclassification, and put forward the concept of cost early-warning. Through some deep study of Bayes Discriminant Analysis, judge this theory has conditions of cost early-warning model. Using simulated sample of100; build the general discriminant analysis without considering misclassification cost, and Bayes discriminant analysis considering the different misclassification cost by setting the decision matrix. Furthermore find out that Bayes discriminat analysis has a smaller misclassification loss or average misclassification loss, and also a slight improvement on overall prediction accuracy. Bayes discriminant analysis directly chooses an evaluation standard according to the minimization of ECM, and deduces the discriminant criterion; and the general discriminant analysis gets the misclassification rate of each type on the basis of minimum TMP, and then transform if to average misclassification loss. It is considered that Bayes discriminant is set beforehand, and general discrimination is set in advance, the "thing" in fact refers to the sample, in other words, one is setting ECM in the population before sample selection, and another is setting ECM in the selected sample. These two kinds of thoughts have their own advantage.