Credit Risk Prediction Based On Process Observable Information

Since the birth of finance,risks have been accompanied by it.With the rapid development of online loan platforms,credit risks emerge in an endless stream.In August 2016,the regulatory authorities issued the interim measures on the management of the business activities of online lending information intermediaries,guiding the way for the standardized development of the industry.Financial institutions need to forecast various indicators of credit risk,whether for their own needs of estimating expected losses and gains or for the requirements of regulatory authorities on financial institutions.The Basel II issued in 2004 fully reflects the credit risk of a transaction,which includes the Probability of Default of the debtor and the severity of the loss caused by the default — the Loss Given Default.Most of the current credit risk prediction models take the borrower's characteristics,such as income level and credit score,which can be observed at the initial time,as the input item X,take the final predicted default probability and the Loss Given Default as the output item Y,and then select the appropriate method to establish a supervised learning model that uses the input item to predict the output item.Based on the research of the Probability of Default and the Loss Given Default,this paper innovatively proposes a credit risk research framework based on process observable information,taking the public data of lending club,the largest online lending platform in the United States as the research sample.Traditional credit risk prediction models usually observe and collect historical data to form a data set containing the observed input X and output Y,so that the data set can be used as the training data to complete the training and establishment of the model,but other observable events that may be closely related to X and Y during the process from observation to X to observation to Y are ignored.Observation events can easily lead to incomplete information mining and reduce the accuracy of the algorithm.Therefore,this paper constructs a credit risk research framework based on process observable information and proposes a stratification model and a mixture model.Compared with traditional forecasting a single model of credit risk,this paper put forward the credit risk based on process information to observable research framework for numeric data type or category data type prediction compared with single forecast model can been improved largely,and for the risk assessment for the borrower credit risk areas,improving the risk control system has important theoretical significance and practical significance.