Twitter Sentiment Mining for Cryptocurrencies with Deep Learning Approach

Due to their volatility in price, the cryptocurrencies are used widely in different financial and business undertak-ings. Advancement in the processing of natural language and giant data analytics has caused the inception of automated systems that assess sentiments in online communities, which have now been crucial in facilitating pub-lication of opinions by individuals in this forum. Twitter, one of the most popular social media there are, has sev-eral tweets about cryptocurrencies. With this information, we can predict the shift of bitcoin prices by utilizing deep learning (DL). Studying the sentiment of Twitter, the researchers aim to come to know more and analyze factors that are driving volatile price of cryptocurrencies. The central issue in this article is the lack of the standard model that may be referred to for the analysis of textual emotions, although textual emotions are among the reasons behind the increase and decrease in the price of cryptocurrencies. The aim of this article is to classify the sentiments for an expression into three classes. neutral, negative, and positive. The techniques that have been used include Word embedding Fast-Text model, Long-short-term memory networks (LSTMs), re-current neural networks (RNNs), gated recurrent units (GRUs), onedimensional convolutional neural networks (CONV1D), a bi-LSTM+CONV1D, Recurrent- Convolutional Neural Networks (Three different datasets from Kaggle which contain different tweets on cryptocurrencies have been used in this research. The Long Short-Term Memory (LSTM) generated the best results in comparison with other similar techniques and outperformed other approaches in determining the sentiment conveyed within the text of the cryptocurrencies.