Stock price prediction using linear regression based on sentiment analysis

Abstract

Stock price prediction is a difficult task, since it very depending on the demand of the stock, and there is no certain variable that can precisely predict the demand of one stock each day. However, Efficient Market Hypothesis (EMH) said that stock price also depends on new information significantly. One of many information sources is people's opinion in social media. People's opinion about products from certain companies may determine the company's reputation and thus affecting people's decision to buy the stock of the company. When using opinion as primary data, it is necessary to make a suitable analysis of it. A famous example using opinion as data is sentiment analysis. Sentiment analysis is a process to determine emotion/feeling within people opinion about something, in this case products of some companies. There are some researches about sentiment analysis used to predict the stock prices. Bollen on his research concludes that people opinion on social media such as Twitter can predict DJIA value with 87.6% accuracy. This shows that there is a relation between sentiment analysis and stock prices. Our purpose on this research is to predict the Indonesian stock market using simple sentiment analysis. Naive Bayes and Random Forest algorithm are used to classify tweet to calculate sentiment regarding a company. The results of sentiment analysis are used to predict the company stock price. We use linear regression method to build the prediction model. Our experiment shows that prediction models using previous stock price and hybrid feature as predictor gives the best prediction with 0.9989 and 0.9983 coefficient of determination.