Direction prediction based on sentiment analysis in Bitcoinusing deep learning algorithms

Emotions form a very important and fundamental aspect of our lives. What we do and say reflects some of our feelings in some way, though not directly. To understand the basic behavior of a person, it needs to be analyzed through some emotional data, also called affect data. This data can be text, voice, facial expressions, etc. it could be. With the advent of social networking sites, many people have turned to analyzing the content found on these various sites. Twitter is one of these social networking sites. Positive, negative or neutral emotional values are revealed through people's feelings and thoughts about a subject. Doing sentiment analysis on Twitter is a very important and challenging task. In this thesis, an overview of Bitcoin-related sentiment and its impact on bitcoin value is presented by utilizing the power of deep learning architectures and machine learning methods. English tweets shared on Twitter between 12 December 2021 and 13 March 2022 were collected to be used within the scope of the thesis. People's feelings about Bitcoin were evaluated using TextBlob, a natural language processing (Natural Language Processing, DDI) tool. Then, GloVe-based BiLSTM model is proposed for emotion classification within the scope of the thesis. The performance of the proposed model is tested separately with TF-IDF and GloVe word embedding approaches using basic machine learning algorithms and CNN, LSTM deep learning architectures. Experimental results prove the success of the proposed model.