Stock Market Movement Prediction using Disparate Text Features with Machine Learning

Abstract

Forecasting stock market movement is a widely researched topic both in academia and industry. Accurate forecast of stock direction can help investors to acquire opportunities for gaining profit in the stock exchange. Predicting stock market due to its dynamic, non-linear and complex nature is inherently difficult. One of the weaknesses of existing stock movement prediction research is that using only sentiment-based features extracted from social media do not completely harness underlying stock behaviour.Finding out which factors are the most significant presents a monumental challenge. Thus, in this research, we will integrate several factors that can affect the stock prices by integrating sentiment analysis with important textual features with relevant lags with the aim to construct more reliable and realistic sentiment representation. To evaluate the performance of our approach, we present a case study based on the AMZN NASDAQ stocks. The experiment results show that random forest model with important features was able to predict the AMZN stock movement direction and to outperform other prediction methods.