Radzicki, Michael J.
For this project, we explored the use of text mining, clustering, and machine learning models to develop a system that combines technical and sentiment analysis to determine the movement of a stock. The final result of our project is a system comprised of a novel sentiment analysis used as input for the larger recurrent neural networks, each trained on a cluster of stocks from the S&P 100. Experimental results show that our system can predict upward movements in stock price over a 65-minute period with up to 77% accuracy for a specific cluster compared to 52% of randomly guessing for the same cluster.
Worcester Polytechnic Institute
Major Qualifying Project
All authors have granted to WPI a nonexclusive royalty-free license to distribute copies of the work, subject to other agreements. Copyright is held by the author or authors, with all rights reserved, unless otherwise noted.
Social Science and Policy Studies