Estimation of Stock Price Using LSTM Algorithm and Sentiment Analysis

Predicting price of the stock is a complex job and it is almost impossible to predict the exact amount of the stock. However, it’s possible to estimate future amount of the stock through searching at historical information. There are various machine learning models available with which stock price can be estimated. In this paper, initially a detailed survey is presented that is based on a comparison and examination of various deep learning algorithms for estimation of stock price. It includes Linear Regression, Support Vector Machine, Auto Regressive Integrated Moving Average, Convolutional Neural Network, Long Short-Term Memory, Artificial Neural network. It is observed that the LSTM algorithm has a better performance as compared to the others. In this work a univariate model is trained to estimate the price of the stock using LSTM model. To improve the results further a sentiment analysis based on twitter data is performed and a multivariate LSTM model is trained. The results presented in this work are based on three stocks Britannia, HDFC bank and Axis bank. The state-of-the-art natural language processing tool BERT are used to recognize the sentiments of text, and the long short term memory neural network (LSTM), which is good at analyzing time series data, is applied to forecast the stock price with stock historical transaction information and text sentiments The various performance parameters are also discussed in this paper. The analysis is performed based on changing number of epochs and it is observed that error percentage reduced with increasing number of epochs.