Macroeconomic indicators and machine learning for financial bubble detection in the Indian stock market

This study investigates financial bubbles in the Indian stock market using the NIFTY 50 index and macroeconomic indicators. Employing the Backward Sup Augmented Dickey–Fuller test, we identified significant bubble occurrences in 2011, 2014, 2016, 2018, 2020 and 2021. Subsequently, supervised machine learning models, including SVM, Voting Classifier and XGBoost, enhanced by PCA for dimensionality reduction and ADASYN for class balancing, were developed to predict future bubbles, achieving a 96% accuracy and an F1-score of 0.98. While these models demonstrate impressive predictive power, understanding the psychological factors driving investor behavior remains crucial. This research underscores the importance of integrating machine learning with insights into human psychology for effective financial decision-making. The use of PCA with ADASYN in conjunction with ensemble learning (Voting Classifier) is a novel approach for bubble detection in the Indian market context.