This project focused on predicting market volatility regimes by building and comparing machine learning models (Random Forest, Logistic Regression, KNN) trained on 20 years of SPY historical data and derived technical indicators. After feature engineering and hyperparameter optimization, the Random Forest model demonstrated a good ability to classify high vs. low volatility days, achieving ~66% accuracy and identifying key predictors like VIX and historical volatility through feature importance analysis. While Logistic Regression scored slightly higher on accuracy, the Random Forest model proved effective overall for this classification task.