Can machines learn Finance? This is the title of a recent paper by Israel, Kelly, Moskowitz (2020) where they discuss the premises of machine learning (ML) techniques for predicting asset returns. Gu, Kelly, Xiu, (2020) provide empirical evidence for tackling this question. They apply a set of different machine learning methods for predicting stock returns and they find that they are not only successful but they also outperform standard linear methods used in the financial literature.

Machine learning methods allow for more flexibility in capturing the relation between predictors and expected returns. Contrary to linear methods, ML can easily leverage on the non-linearities and interactions between the predictors. On the other side, ML models, because of the many degrees of freedom, are more prone to fit noise, especially in a low-signal to noise environment which characterizes financial markets.

During these challenging projects, you will apply advanced statistical techniques to large datasets containing for example individual company characteristics or macro-economic series. A good understanding of statistics and machine learning, a practical mindset and the ability to work with large amounts of data are crucial for this project. On top of that, you have strong programming skills, preferably in Python.

The goal of the project is to generate insights that will help Robeco to better translate our alpha predictions into solutions accustomed perfectly to clients’ risk profiles.

Examples of previous internship projects

• Statistical clustering techniques in stock and bond selection

• Predicting company earnings using Machine Learning