Predicting Stock Outperformance by Machine Learning

Automated Stock Picking using Random Forests

AuthorsChristian Breitung

Paper linkhttps://papers.ssrn.com/sol3/papers.cfm?abstract_id=3978532

A global stock ranking strategy using a non-refitted random forest classification model evaluates outperformance probabilities, constructing long-short portfolios with high returns, lower volatility, and improved Sharpe ratios across liquid stocks.

Machine learning captures non-linear stock patterns and anomalies, enabling better stock selection than traditional models. Random forest classification provides risk-adjusted portfolios that outperform in out-of-sample tests, aiding informed investment decisions.