Wars, trade disputes, and diplomatic crises are having a bigger and bigger effect on global financial markets. Developing countries like India are especially vulnerable to these outside shocks since they are linked to global financial flows. This study compares the prediction capabilities of three Machine learning models namely, classical statistical models (ARIMA), XGBoost, and Long Short-Term Memory (LSTM) networks while predicting how geopolitical risk (GPR) may affect Indian stock market indices like the NIFTY 50 and SENSEX. RMSE, MAE, and R2 are used to rate the models. Research shows that ML-based time-series models, especially LSTM, perform better than conventional models and are able to capture non-linear dependencies and market responses to geopolitical shocks. For various investors, the portfolio supervisors and research advisers, this method helps them with various useful information that can make the results more accurate of the risk taken.