Comparison of Machine Learning Algorithm for Stock Price Prediction Using Sentiment Analysis

Abstract

Forecasting companies' stock market prices are one the interesting topics for analysts and researchers. Although a company's stock price can be unpredictable, long-term forecasts can help but it is dependent on many factors such as the company's business model, change in leadership, and investors' mood. It has been found to be insufficient to predict stock values just on the basis of historical data or textual information. Previous research in sentiment analysis have shown a strong correlation between the movement of stock prices and the publication of news stories. At different levels, a number of sentiment analysis research have been attempted utilizing methods. In this paper, we made a comparison of various Machine Learning methods on five datasets of financial news related to the company and domains in which the company. Encouraging results are obtained using 13 models i.e., Linear Regression, Ridge Regression, Lasso Regression, Random Forest, Naive Bayes, Logistic Regression, LSTM, ARIMA, Logistic Regression, Support Vector Machines, Decision Tree, Boosted Tree, and ensemble method which depict polarity of news articles being positive or negative and the accuracies are 93.90%, 92.31 %, 92.27%, 85.44%, 84.65%, 84.65%, 94.73%, 90.13%, 82%, 83%, 72%, 70%, 95.11 % respectively.