Feature selection of nonperforming loans in Chinese commercial banks

Abstract

In recent years, huge amounts of nonperforming loans (NPLs) of commercial banks have become one of the biggest obstacles constraining reform and development in Chinese commercial banks. Finding a way to control the banks' NPLs is a core issue that it continues to be explored and researched in the finance. In this paper, PCA and relief algorithm in data mining methods were adopted to extract and analyze NPLs characteristics in commercial banks through contrasting the performing and nonperforming loans records, based on the predecessors' literatures. In this paper, a bank's loans data with 96 features and 10415 samples is collected. At last, we construct nonperforming loans of commercial banks classification model. Our research is very important for capturing warning signal timely, detection of NPLs and sound operation of commercial banks.