فهرست نشریات علمی دسترسی آزاد ایران

Financial markets are characterized by volatility and non-stationarity, posing challenges for effective investment strategies. Traditional approaches, such as trend-following and Momentum (MOM) strategies, have long been employed but often yield inconsistent results across stock markets and under varying market conditions. With the rise of Machine Learning (ML), data-driven methods offer new opportunities for enhancing trading decisions; however, their comparative effectiveness against traditional approaches remains underexplored. This study presents a sector-specific analysis of trading strategies within the Consumer Discretionary, Healthcare, and Information Technology sectors of the S&P 500. Structured in two phases, Phase 1 evaluates the performance of traditional strategies—trend-following, Momentum (MOM), and buy-and-hold—using risk-return metrics such as annualized return, volatility, and the Sharpe e ratio. Phase 2 applies ten ML models, including ensemble methods, to develop enhanced strategies based on technical indicators. Models are assessed on both predictive accuracy (accuracy, precision, recall, and F1-score) and financial performance through a 10-year backtest. Findings suggest that ML-enhanced strategies can outperform traditional approaches, particularly in terms of risk-adjusted returns. The originality of this study lies in its sector-based framework, which not only compares conventional and ML-driven strategies but also evaluates them using a dual lens of predictive accuracy and risk–return metrics. Furthermore, for each sector with distinct characteristics, the study identifies the ML models that achieve the best balance of accuracy and financial performance.