Sentiment Analysis of Stocktwits Data With Word Vector and Gated Recurrent Unit

Prediction of stock movements is important in the business world for knowing the movement of stock both for buying and selling goods. Stock is a financial product characterized by high risk, high return and flexible trading, which is favored by many investors. Investors can get abundant returns by accurately estimating stock price trend. Historical price is often used to predict the stockprice, it can only estimate the periodical trends of the stockprice. However, there could be a particular event that may affect the price. So it cannot capture sudden unexpected events. Social media texts like tweets can have huge impacts on the stock market. By analysing the sentiments of social media information, unexpected behaviour of the price trend could be detected. In this study, we propose to use Gated Recurrent Unit (GRU) for predicting the sentiment of tweets related to stockprice. We implement word vector, in particular word2vec, as features for GRU. Our experiments show that the proposed method is better than other deep learning based sentiment analysis such as BERT (Bidirectional Encoder Representations from Transformers)  and BiLSTM (Bidirectional Long Short Term Memory).