Tweet Based Sentiment Analysis for Stock Price Prediction

Abstract

Economic, social and political variables have a large influence on the stock market. Stock markets can be influenced and driven by a change of variables, both internal and external. Stock prices alter from moment to moment due to variations in supply and demand. Various data mining methods are routinely used to overcome this challenge. The subject of sentiment analysis for rising and falling stock prices is addressed by analyzing a number of Deep Learning (DL) techniques. Older price prediction methods could estimate prices in real-time, but be short of the information and corrections to predict price deviations. In this work, cryptocurrency sentiment analysis is commonly used to predict cryptocurrency market prices. Long short-term memory (LSTM) and Bidirectional Encoder Representation from Transformer (BERT) are two models fused together to frame the proposed Chalk Enhanced Algorithm (CEA) model to increase the effectiveness of analyzing stock market ups and downs. CEA is the ensemble model of LSTM and Bert. This model overcomes overfit problems and uses less time. Comparative analysis of LSTM, BERT and proposed CEA models are carried out using four performance metrics (accuracy, precision, recall, and F1-score), the proposed CEA model is the best model for gauging stock sentiment analysis. The higher value of accuracy, precision in the proposed model proves that this novel method is more effective than the existing one. This allows both new and existing investors to invest with confidence and take advantage of new opportunities.