A Stock Prediction System Based on News and Twitter

Abstract

As more concerns by the public about stock markets grow bigger, the more the people’s attention is drawn to a systematic method to predict stock prices that fluctuate. More importantly, as the modern stock markets react very sensitively to information for their stock prices, it is very important to predict the prices for investors. For that, this study shall utilize opinion mining and mechanical learning, which are widely used to analyze the meaning of information in systematic ways on analyzing data from news and Twitter to suggest a system that predicts stock prices. The stock price prediction system consists of a data collector, vocabulary analyzer, sentiment analyzer and stock price predictor. The stock price predicting steps consist of collecting contents of news and Twitter, extracting vocabularies by using morpheme analysis, executing sentiment analysis then predicting stock prices via mechanical learning. In order to evaluate the usefulness of the suggested method, we used the stock data for the last whole year on 7 companies in the bio industry that are most sensitive to information for the tests, and the accuracy of the results showed above 80%. The results of this study can be regarded as one of the methods to effectively predict stock prices of companies from various backgrounds in this modern information era that changes dramatically every moment.