AI-Driven Stock Market Prediction: Can Machine Learning Improve Portfolio Performance?
0 viewsAbstract
This study investigates the comparative effectiveness of machine learning models against conventional financial portfolio optimization techniques. While traditional models like Markowitz’s Modern Portfolio Theory rely on linear assumptions and normal distributions, modern financial markets are characterized by non-linear patterns and structural breaks that often elude classical econometric methods. Using historical stock market data across diverse industries, this research employs an empirical framework to compare the performance of Mean-Variance Optimization against predictive portfolios driven by Random Forest, Gradient Boosting, and Neural Network algorithms. Performance is evaluated using key financial metrics, including the Sharpe Ratio, average returns, and portfolio volatility. The results indicate that machine learning models generally outperform conventional approaches in short-term return forecasting by identifying complex, non-linear relationships within the data. AI-driven portfolios demonstrated superior risk-adjusted returns during stable periods. However, the study also reveals that machine learning model performance can be inconsistent during periods of high market volatility and remains susceptible to overfitting. The research concludes that while machine learning is not a perfect solution, it represents a significant evolutionary step in financial modeling, serving to enhance rather than replace the predictive accuracy and optimization of modern investment strategies.
Keywords
Artificial Intelligence in Finance, Efficient Market Hypothesis, Financial Forecasting, Gradient Boosting, Machine Learning, Mean-Variance Optimization, Modern Portfolio Theory, Portfolio Optimization, Random Forest, Sharpe Ratio