Portfolio selection based on predictive approach

Abstract

The paper analyzes a conceptual value investor behavior that uses stock fundamentals to predict the best-performing portfolio according to the knowledge it has at its disposal from previous periods. A class of machine learning algorithms based on inductive learning is used to learn from past data and predict by induced rules which stock to select for the final portfolio. This approach has led not only to corroborate with other research on the relevance of financial fundamentals to explain the future returns of stock but also to design a strategy that generates significant excess returns compared to the market index, S&P 500.