A decision tree classifier for credit assessment problems in big data environments

Abstract

Financial institutions have long sought to reduce the risk of consumer loans by improving their credit assessment methods. As new information and network technologies enable massive data collections from many different sources, credit assessment has become a challenge in the big data environment. Complicated processing is required to deal with vast, messy data sources and ever-changing loan regulations. This study proposes a decision tree credit assessment approach (DTCAA) to solve the credit assessment problem in a big data environment. Decision tree models offer good interpretability and easily understood rules, with competitive performance capabilities. In addition, DTCAA features various data consolidation methods to eliminate some of the noise in raw data and facilitate the construction of decision tree. By using a large volume data set from one of the biggest car collateral loan companies in Taiwan, this study verifies the efficiency and validity of DTCAA. The results indicate that DTCAA is competitive in various situations and across multiple factors, in support of the applicability of DTCAA to credit assessment practices.