Path Signatures in Machine Learning-based Analysis

This paper examines the application of the rough paths theory in modelling of financial time series. The theory of rough paths provides a way to effectively and efficiently capture the relevant information about rough signals, which can be used in machine learning modelling. This approach is applied to twelve stock market indexes with a goal to predict the sign of their daily returns (positive or negative) and their realized daily volatility.

Autorentext

Completed Bachelor's degree studies at the University of St. Gallen, with major in Economics. Doing research in the application of machine learning to modelling of financial markets. Currently pursuing the Master's degree program MSc Statistics at the ETH Zurich.