An improved gradient boosting tree algorithm for financial risk management

Abstract

Financial risk assessment (FRA) is an essential process in financial institutions determining a company’s creditworthiness. This paper introduces a new wrapper feature selection with a clustering-based FRA model to assess the financial status. This study involves three different phases of operations such as feature selection, clustering, and classification. The proposed model initially designs an Information Gain Directed Feature Selection algorithm that offers to rank to the features utilising the information gain. In addition, the proposed model also involves an improved K-means clustering technique to cluster the data. Finally, the gradient boosting tree classifier model is executed to perform the classification process. The proposed model tested using two benchmark datasets. The simulation results indicate that the projected FRA model obtains maximum accuracy values of 95.68% and 94.76% on the applied datasets.