Can Memes Beat the Market? Forecasting Financial Asset Returns by Using Social Media Data and Machine Learning

The objective of the research presented in this paper is to demonstrate the effect of the social media data on the return on investment in crypto trading. To do that, three methods were used in vectorized back-testing on selected 17 crypto-currencies, namely, Machine Learning-driven strategy, sentiment signal or feature-driven strategies and hybrid strategies. Two sentiment signal strategies are trialed, so-called naïve sentiment momentum and percentile strategy. One hybrid strategy was tested, namely, combined crossover and naïve sentiment momentum strategy. The performance of all methods is assessed by using normalized ROI and Sharpe ratio as indicators. The best results in back-testing on equally weighted portfolio were achieved by using ML-driven strategy, namely normalized ROI of 6.14 and Sharpe ratio of 1.98.