Stock Selection and Portfolio Optimization: A Machine Learning Approach

Details

Supervisors

Vrins, Frédéric

Faculty

Louvain School of Management

Degree label

Master [120] : ingénieur de gestion, à finalité spécialisée

Abstract

A challenging problem in active asset management is the selection of stocks and their weighting to form a balanced portfolio. In this regard, we employ machine learning techniques that aim to predict which stocks will outperform their benchmark during the coming quarter. We then integrate these picks into a portfolio in such a way as to minimize its variance. Our approach is evaluated using a rolling window that allows us to measure the performance over an out-of-sample period from August 2004 to May 2021. Our empirical findings show that machine learning-based selection models can be used to construct portfolios that exhibit a higher return and a lower risk than the benchmark. The best performing model obtained a Sharpe ratio of 0.811 while the benchmark index had one of 0.681.