Towards development of robust machine learning model for Malaysian corporation: a systematic review of essential aspects for corporate credit risk assessment / Zulkifli Halim and Shuhaida Mohamed Shuhidan

This study aims to review the essential aspects of credit risk assessment. The scope of this study is the credit risk assessment studies that used machine learning or used Malaysian data. This study is an overview of the development of robust machine learning for Malaysian corporation credit risk assessment. This study used a systematic review as the methodology. After thorough searching, this study has selected 20 studies to be reviewed. As a result, three essential aspects are identified: the variables, the features, and the methods used for financial distress prediction. This study found that financial ratios, macroeconomics, and corporate governance indicators are essential in credit risk assessment studies. The debt ratio was recorded as the most widely used ratio, found in 14 studies, followed by the liquidity ratio, used in 12 studies. In addition, the studies performed using Malaysian data show that the debt ratio and liquidity ratio are significant. Support vector machine (SVM) and genetic algorithm (GA) are among the best methods to be used. Recurrent neural network (RNN) is the latest credit risk assessment method to solve the time series data problem. In conclusion, all the essential aspects identified in this study should be considered in any credit risk assessment study.