Tools
Tools
Toriola, Adebayo J. (2021) Prediction of Bitcoin Prices Using Deep learning and Sentiment Analysis Based on Bitcoin Tweets. Masters thesis, Dublin, National College of Ireland.
Preview |
PDF (Master of Science)
Download (3MB) | Preview |
Preview |
PDF (Configuration manual)
Download (3MB) | Preview |
Abstract
Price fluctuation and volatility of Bitcoin currency in the last few years has caught the attention of many researchers. Many researches are still ongoing to identify and define the sum of exact factors that influences Bitcoin price. Speculation expressed on sentiments has been identified by many researchers as a significant factor of the price volatility. Twitter is one of such platforms where users express their sentiment on issues. However, these data is a time series because it is recorded at regular interval, hence it is necessary to look at the data overtime to predict the next value therefore, in this research VADER sentiment analyser is used to analyse and assigned sentiment scores to Bitcoin-related tweets, the score is merged with historical price. Thereafter, ARIMA and LSTM models were applied to analyze the merged data in order to predict the price movement. Time series analysis is performed on the merged data and it reveals that there is a positive correlation between the Twitter sentiment and the bitcoin price. Finally, the execution time of these two models were evaluated on both local machine and cloud environment and the LSTM model achieve a good RMSE of 0.014 within 1.09 minutes for per minutes data on GPU and and RMSE of 0.018 within 1.29 minutes for for the per hour data.
Actions (login required)
 |
View Item |