Portfolio Construction with Crowd Intelligence: A Machine Learning Approach

Details

Supervisors

Vrins, Frédéric

Faculty

Louvain School of Management

Degree label

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

Abstract

In this era where users increasingly share their opinions on the internet, this study assessed the relevance of crowd intelligence in the construction of portfolios. The data used was composed of opinions about stocks’ future returns between April 2006 and April 2022. Machine learning algorithms, namely linear regression, artificial neural networks and extreme gradient boosting, were used to predict the expected return of the opinions. These predictions were combined with an equal-, constrained equal-, and an expected return to historical standard deviation weight attribution method to produce a set of portfolios. The portfolios were applied to a one-day and 61-day holding period. The best performing portfolio with the one-day holding period obtained a Sharpe ratio of 0.899 and an APY of 28.22%, while the benchmarks had a Sharpe ratio of 0.67 and 0.576, and an APY of 13.72% and 10.54%. For the 61-day holding period, the best portfolio obtained a Sharpe ratio of 3.79 and an APY of 9.95%. Overall, the empirical findings showed that machine learning predictions based on crowd opinions can be used to construct portfolios that exhibit a higher return and/or higher risk-adjusted returns than the benchmarks. To conclude, the portfolio performances showed encouraging results that crowd intelligence is performant in portfolio construction.

Citations

• APA

Morelle, H. (2022). Portfolio Construction with Crowd Intelligence: A Machine Learning Approach. https://dial-mem.test.bib.ucl.ac.be/handle/2078.2/30168