The Bank's Corporate Loan Default Risk Forecast Analysis

How to precisely identify corporate loans with potential default risk besides protect the privacy of data is a popular topic for the banking industry and academia. Distribution of a serious imbalance between information and data for business loans, we propose a approach to select the optimal classification method of forecasting model.The proposed method is divided into two parts:pre-processing and data analysis. Pretreatment is worked by the bank itself, including deleting some information which is not related with the experiment, filling the missing values and independent component analysis and so on. Why do this is in order to prevent the disclosure of some sensitive core data, thus causing heavy losses to banks and customers. The data analysis part is specified by the banking institutions, including over-sampling for unbalanced data classification algorithm selection, classification and evaluation indicators selection, multi-objective decision method to select steps, the part from the processed data useful feedback to the bank, to help banks distinguish loan defaults in accuracy, reducing the risk of bank loans, to improve the bank’s income.In our experiment,Business loan data to a Chinese state-owned bank Sichuan branch, for example.We get in the weighted case, the sort of optimal multi-objective decision making method TOPSIS is better than other optimal multi-objective decision making methods.