Assessing the Efficient Market Hypothesis for Cryptocurrencies with High-Frequency Data Using Time Series Classification