Predicting of Banking Stability Using Machine Learning Technique of Random Forests

Abstract

The purpose of this research is to create a predicting model of banking stability in Indonesia. Authors use a small set of explanatory indicators primarily related to the banking industry and some relevant economic variables. Among the indicators candidate to be used in this study are the indicator of banking industry, the money markets, capital markets and creditors, and the macro-economic indicator. The source of data in this research is obtained from CEIC and SEKI (Indonesian Economic and Financial Statistics) that published by Central Bank of Indonesia from 2004 and 2011. Principal Component Analysis is used to avoid the multi-collinearity problem when construct the model. Authors train the model using Random Forest Regression with data over the 2004-2007 period, and made predictions of global financial crisis that happened in 2008/9. Python 2.7.10 and scikit-learn version 0.20.0 module has been exploited for simulations and evaluation of the model. Numerical illustration is provided to demonstrate the efficiency of proposed model. As the result nine most components analysis obtained as input for the machine learning model with explained variance ratio around 97%, accuracy around 89%, and precision 91% and mean absolute error around 11%.