Implementing machine learning in the stock picking process of Nova students portfolio

Abstract

In a time when algorithmic trading accounts for over 50% of US equities’ traded volume, this work project proposes a holistic approach to the implementation of Machine Learning in the Stock Picking process of the Nova Students Portfolio. The presented algorithms can help investors in the identification of the features that drive stock returns and results show that our predictive algorithm provides an edge in the selection of outperforming stocks. An investor using our method from 2006 to 2019 would have achieved an annualized return of 4.8% in excess of the S&P 500 and an Info Sharpe gain of 0.2.