An Efficient Method to Predict the Tata- Motors Stock Price using Hybrid Machine Learning Methods

Abstract

Stock market analysis has always been an important aspect of every country's financial sector. As of now, various research has been done to predict the stock market prices but only considering the technical stock data. However, the problem lies in combining the technical data of stock prices and news sentiments from financial news data so that prediction can be done with much greater accuracy. In our paper, we have designed a stock price prediction system and proposed an approach in which technical stock Data is evaluated by technical means and news sentiment data is represented in the form of sentiment vectors using sentiment analysis. We have deployed Particle Swarm Optimization (PSO) to tune the hyper- parameters of the Support Vector Machine for regression (SVR), thus providing better results. We have done experiments on the Tata Motors stock price data and compared our approach with [1] who have deployed the SVM-PSO model with basic technical features taken into consideration. Our model SVR-PSO with financial news data gives a Mean Absolute Percentage Error of 0.29% as compared to the standard SVM- PSO which gives a Mean Absolute Percentage Error of 0.71 %