Social media has experienced rapid growth since its introduction. This includes providing platforms through which people share and collect news and opinions in real time. This thesis explores whether the social media platform Twitter can be used as an information source to predict the future stock price of a company. We built a system that predicts stock prices for Google, Apple, and Yahoo based on a set of tweets collected in real time. The system has two major components. Advanced Natural Language Processing techniques, including Sentiment Analysis, are employed to determine the polarity (positive or negative) of each tweet. Using a feature set created from these tweets as input, Support Vector Regression is used to make discrete future stock price predictions. Various experiments were conducted to find the best configuration of the datasets and to optimize the prediction accuracy. The results are promising, but further investigation is necessary for the practical usage of the system.


Gray, Simon

Second Advisor

Pierce, Pamela


Computer Science; Mathematics


Artificial Intelligence and Robotics | Operational Research | Statistical Theory

Publication Date


Degree Granted

Bachelor of Arts

Document Type

Senior Independent Study Thesis



© Copyright 2014 Kazuki Kyotani