In recent years, social media has become ubiquitous in social networking and content sharing. Yet the content that is generated from these websites remains largely untapped. This Independent Study investigates the usability of Twitter data for predicting trends of share price directionality in stock markets. In particular, it uses Tweets about Tesla and Boeing together with five learning algorithms (two of them being neural networks) to predict these trends.
First, the data is pre-processed and analyzed for sentiment (i.e. each Tweet is labeled as positive or negative). Then, these labeled Tweets are used together with the actual stock prices in the five learning methods. The best accuracy is obtained using a cascading neural network with a windowing technique that incorporates the share prices and the sentiment data from the five days prior to the day being predicted. The researchers found a moderate correlation (62\% correct prediction) between the model's predictions of directionality and the true directionality of the stocks sampled. Better results could be achieved with better sentiment analysis and with larger datasets.
Catlin, Tyler, "Predicting Stock Market Directionality Using Twitter and Neural Networks" (2016). Senior Independent Study Theses. Paper 7221.
Artificial Intelligence and Robotics | Portfolio and Security Analysis
Bachelor of Arts
Senior Independent Study Thesis
© Copyright 2016 Tyler Catlin