Accurately identifying which sectors will outperform or underperform is key for active asset allocation. This study introduces a three-class classification framework–Top, Mid, Bottom–to rank monthly sector performance in the Tunisian stock market. Instead of predicting raw returns, we model relative sector ranks using a hybrid set of market-based technical indicators and macroeconomic variables. We evaluate three machine learning algorithms (Random Forest, XGBoost, and Multinomial Logistic Regression) across different input combinations: macro-only, market-only, and combined features. The combined models achieve strong performance, with accuracy reaching up to 99% for logistic regression and robust F1-scores. Results confirm that momentum, volatility, and recent returns are key short-term predictors, while macroeconomic variables add complementary value. Our findings suggest that rank-based classification can effectively support sector rotation strategies, offering a practical decision tool for investors in emerging markets such as Tunisia.