Artificial Intelligence-Driven Financial Risk Prediction and Portfolio Optimization: A Machine Learning Approach from the Indian Equity Market

As financial markets become more complex, data-driven risk identification and portfolio management becomes the focus for the future. In this study we have developed an integrated machine learning (ML) framework to predict financial risk and portfolio optimization in the Indian stock market. Using 32 NSE-listed large-cap securities in eight sectors during 2021–2025, the study constructs 14 key financial risk profiles based on volatility, momentum, drawdown, technical indicators, and macro-financial variables in India. ML models are used to classify forward 20-trading-day high-risk and forecast 20-day returns. Gradient Boosting and Neural Network classifiers perform best on the scale of 0.794 and 0.798 and F1-scores of 0.830 and 0.828 respectively. However, the regression models fail to outperform the naïve mean forecast, indicating that short-term returns are not yet well predicted. A portfolio comparison between equal-weighted, Markowitz mean-variance and ML-risk-tilted strategies shows that the Markowitz portfolio achieves the best risk-adjusted performance with a Sharpe ratio of 3.84 and Sortino ratio of 10.52. There is no significant difference in mean returns across sectors, while India VIX shows a weak positive relationship with realized volatility. The results show that ML can be a good early warning system on market risk and classical portfolio optimization is still a useful tool to construct a portfolio. The study therefore recommends the use of ML risk signals as a complementary tool to improve, instead of replace, traditional portfolio management strategies.