Research Article

Anticipating Cryptocurrency Prices Using Machine Learning

Figure 14

Daily geometric mean return obtained under transaction fees of . The geometric mean return computed between time "start" and "end" using the Sharpe ratio optimisation for the baseline (a), Method 1 (b), Method 2 (c), and Method 3 (d). Note that, for visualization purposes, the figure shows the translated geometric mean return G-1. Shades of red refer to negative returns and shades of blue to positive ones (see colour bar).