A K-Means Classification and Entropy Pooling Portfolio Strategy for Small and Large Capitalization Cryptocurrencies

Abstract

In this study, we propose three portfolio strategies: allocation based on the normality assumption, the skewed-Student t distribution, and the entropy pooling (EP) method for 14 small- and large-capitalization (cap) cryptocurrencies. We categorize our portfolios into three groups: portfolio 1, consisting of three large-cap cryptocurrencies and four small-cap cryptocurrencies from various K-means classification clusters; and portfolios 2 and 3, consisting of seven small-cap and seven large-cap cryptocurrencies, respectively. Then, we investigate the performance of the proposed strategies on these portfolios by performing a backtest during a crypto market crash. Our backtesting covers April 2022 to October 2022, when many cryptocurrencies experienced significant losses. Our results indicate that the wealth progression under the normality assumption exceeds that of the other two strategies, though they all exhibit losses in terms of final wealth. In addition, we found that portfolio 3 is the best-performing portfolio in terms of wealth progression and performance measures, followed by portfolios 1 and 2, respectively. Hence, our results suggest that investors will benefit from investing in a portfolio consisting of large-cap cryptocurrencies. In other words, it may be safer to invest in large-cap cryptocurrencies than in small-cap cryptocurrencies. Moreover, our results indicate that adding large- and small-cap cryptocurrencies to a portfolio could improve the diversification benefit and risk-adjusted returns. Therefore, while cryptocurrencies may offer potentially high returns and diversification benefits in a portfolio, investors should be aware of the risks and carefully consider their investment objectives and risk tolerance before investing in them.

A Comprehensive Investigation of Lane-Changing Risk Recognition Framework of Multi-Vehicle Type Considering Key Features Based on Vehicles’ Trajectory Data

“…To obtain the LC risk level, this paper utilized the K-means algorithm to label the level of LC risk. This algorithm is commonly used to process two-dimensional data, and the basic principle is to find a division scheme of K clusters that minimizes the loss function corresponding to the clustering result iteratively [45]. Additionally, K-means have been proven effective in LC risk clustering by Chen [14] and Wang [35].…”