An evaluation of logistic regression and random forest model as early warning system models for assessing an equity market crisis in ASEAN-5 + 3 countries

To mitigate the impact of the crisis on larger and more liquid markets such as the banking and currency sector, this study focused on targeting a market that initially reacts to unforeseen events - the equity markets. This study aimed to evaluate the predictive ability of Logistic Regression and Random Forest Model as early warning system models, and the key indicators influencing a near-term equity market crisis in ASEAN-5 + 3 countries. The study employed the Confusion Matrix, ROC Curve, and Gini Index to examine the performance of the models using a monthly sample period from January 2005 to December 2022. The empirical results showed that the most occurring key indicators that influence an equity market crisis were Consumer Confidence Index (CCI), Real Effective Exchange Rate (REER), and the S&P 500. Moreover, the findings also showed that the Random Forest Model significantly outperformed Logistic Regression in all performance measures. This study contributed to early warning systems literature in several ways as it first addressed the limited studies that examined the equity market crises on its own and in the ASEAN context. Second, this study addressed the need for a hybrid model capable enough to capture the dynamic nature of financial market crises. Furthermore, this study contributed to the growing body of research on the development of efficient and accurate early warning systems through emerging machine learning and fourth-generational models. Keywords: ASEAN-5 + 3 Countries; Logistic Regression; Random Forest Model; Early Warning System; Equity Market Crisis; Confusion Matrix; ROC Curve; Gini Index