A DATA-DRIVEN QUANTITATIVE FRAMEWORK FOR CRYPTOCURRENCY INVESTMENT PORTFOLIO DESIGN

Abstract

"Cryptocurrencies are digital currencies secured by cryptography techniques, and they use blockchain technology to provide decentralised transactions. Bitcoin, Ethereum, and Ripple are examples of over 37 million unique cryptocurrencies today. Often, cryptocurrencies are traded as assets analogous to stock trading, and this study focuses on such trading and investment opportunities. The main objective of this study is to form optimal portfolios by developing strategies based on risk, return, and price classifications. As the first step, feasible currencies were selected using market information and collected hourly price data for three months. After that, based on the volatility spectrum and the market dynamics, price, risk, and return levels were identified. One of the machine learning techniques, namely K-means clustering, is deployed in this level of identification. Concerning the total portfolio investment, small-scale investments ranging from 0$ - 50$, 50$ - 150$, and 150$ - 500$ are prioritised in the analysis. Addressing the different risk aversion behaviours of investors, strategies such as High Risk-High return, Moderate Risk-Moderate return, Low Risk-High return, and Low Risk-Moderate return have been considered in portfolio formation. A Monte-Carlo simulation framework is formed to investigate the performance of the portfolios consisting of cryptocurrencies belonging to different classes of price, risk, and return levels, to address the above investors’ strategies. The Markowitz optimisation procedure optimises each of the portfolios. On average, the return of the resulting optimum portfolios exceeded the return of random portfolios that have not been formed according to such a quantitative framework. In conclusion, the steps involving price, risk, and currency level identification, strategically forming portfolios, and selecting the optimal portfolio to be exercised, have formed a rigid quantitative framework that can be utilised in cryptocurrency investment to result in better returns. This framework may be used in forming portfolios of any investment level. However, the machine learning algorithms used in the framework must be recalibrated to the cryptocurrency market changes reflected by the price fluctuation as time progresses.