Asset Classification Using Machine Learning Teechniques: The Case of Bitcoin

Kevin Konings & Valentina-Nicoleta Musat

Student thesis: Master thesis

Abstract

We discuss the problem of classifying Bitcoin as a currency or commodity and approach it from a market perspective. Following theory and recent empirical evidence on market dynamics of commodities and exchange rates, we construct a number of features and em- ploy both machine learning technology and classical econometric methods to identify the differences in market dynamics of both asset classes. Our results strongly imply volatility and liquidity as differentiators, as well as show strong predictability of the asset class of a given market independently of the instrument through which it is traded. We finally conclude that Bitcoin is almost universally identified as a commodity, which is in line with past evidence and models, but provides little evidence on theories which consider it as a new asset class with properties of both asset classes.

EducationsMSc in Advanced Economics and Finance, (Graduate Programme) Final Thesis
LanguageEnglish
Publication date2016
Number of pages208
SupervisorsSteffen Brenner