Research on Factor Identification and Contribution of Bank Nonperforming Loans Based on Lightgbm Algorithm

Abstract

In recent years, financial risks and crises have occurred frequently, which has had a far-reaching impact on the world economic pattern. Among them, the bank non-performing loan ratio (NPL) is usually used as a barometer of financial risks, which is used to identify the factors of bank non-performing loans. Combined with the current macro and micro economic data in China, this paper selects 17 indicators from four representative commercial banks from 2010 to 2021 as samples and uses the Light Gradient Boosting Machine (LightGBM) algorithm to empirically analyze the influencing factors of commercial banks' non-performing loan ratio, to quantify the sensitivity of each factor to the non-performing loan ratio. The results show that the influence of the bank's internal finance (0.7874) on the bank's non-performing loan ratio is much greater than the bank's external macro factors (0.2126), and the secondary index with the highest weight in the internal finance category is the medium and long-term loan interest rate (0.3029). Banks should pay attention to endogenous variables, especially the changes in medium and long-term indicators, and timely predict and reduce the risks caused by non-performing loans.