This dissertation was written as a part of the MSc in Data Science at the International Hellenic University, by student Michail Vlachos – Giovanopoulos under the supervision of Prof. Christos Tjortjis.
The rapid growth of social media has significantly increased the value of social media data. Social media platforms have inserted in people’s everyday lives, giving analysts the opportunity to analyze the information and use them to make future predictions for events like cryptocurrency prices, stock market movements, and many more. This project aims at collecting Twitter data about cryptocurrencies, and specifically about Bitcoin, Ethereum, and Dogecoin, conduct sentiment analysis, and perform price predictions based on the sentiment scores. After examining the results of the correlation analysis, it was decided to gather tweets volume and volume of transactions for a longer period, which finally produced very good results. Various machine learning techniques were used in order to further improve the predictions. Finally, the best forecasts were achieved with 91.5%, 93.3%, and 83.6% R-Squared for Bitcoin, Ethereum, and Dogecoin respectively, accompanied by 0.5%, 0.3%, and 0.4% Mean Squared Error for each of the three crypto-currencies respectively.
Collections
Show Collections