Machine Learning enabled Asset Allocation

back to overview

Type and Duration

Preproposal PhD-Thesis, since September 2020


Chair in Finance

Main Research

Wealth Management

Field of Research

Portfolio Management


With increased computing power, more advanced algorithms, and growing data resources, machine learning is widely used in different scientific areas. Machine learning is particularly useful in prediction and clustering tasks and enriches the econometrician's toolbox. Financial machine learning diverges from classical machine learning because financial (market) data suffers from low signal-to-noise ratios, which complicates signal extraction. The high amount of noise also hinders classical asset allocation tasks. Traditional Markowitz mean-variance optimization is unable to outperform a simple 1/N portfolio's return out of sample, which is mainly caused by noise in the data and difficulties in matrix inversion. Machine learning algorithms have the potential to provide financial researchers with new insights. Machine learning based hierarchical clustering already showed promising evidence in this direction. The target of this doctoral thesis will be the further investigation of machine learning's potential in advanced asset allocation. Both clustering and prediction techniques will be applied to evaluate machine learning's strenghts in contrast to classic econometrics.