Stock market analysis using social networks

Nowadays, the use of social media has reached unprecedented levels. Among all social media, with its popular micro-blogging service, Twitter enables users to share short messages in real time about events or express their own opinions. In this paper, we examine the effectiveness of various machine learning techniques on retrieved tweet corpus. A machine learning model is deployed to predict tweet sentiment, as well as gain an insight into the correlation between twitter sentiment and stock prices. Specifically, that correlation is acquired by mining tweets using Twitter's search API and process it for further analysis. To determine tweet sentiment, two types of machine learning techniques are adopted including Naïve Bayes classification and Support vector machines. By evaluating each model, we discover that support vector machine gives higher accuracy through cross validation. After predicting tweet sentiment, we mine historical stock data using Yahoo finance API, while the designed feature matrix for stock market prediction includes positive, negative, neutral and total sentiment score and stock price for each day. In order to capturing the correlation situation between tweet opinions and stock market prices, hence, evaluating the direct correlation between tweet sentiments and stock market prices, the same machine learning algorithm is implemented for conducting our empirical study.