Forecasting S&P 500 sector ETFS returns

: a machine learning approach to sector rotation

Abstract

This thesis develops a machine learning-based dynamic sector rotation strategy, sector ETFs are used as a representation of the United States 11 sectors and of the benchmark S&P500. We use a variety of machine learning models, including LASSO, XGBoost, and Random Forest, to forecast sector returns by utilizing financial market data, market sentiment metrics, currencies and macroeconomic indicators. The top and bottom-performing sectors are chosen to create long, short, and long-short strategies for portfolio construction based on these forecasts. Results show limited predictive power overall, but the strategies built offer modest improvements over the benchmark in certain market conditions.

Date of Award	23 Jun 2025
Original language	English
Awarding Institution	Universidade Católica Portuguesa
Supervisor	Dan Tran (Supervisor)