Joint Analysis of Stock Price Predictive Models Using Past Stock Values and Public Sentiments on Twitter

Abstract

In this study, attempts have been made to predict the stock price movement base on the company Western Digital Corporation. Daily stock price data is used for training and building the predictive models. Four machine learning and deep learning analysis including but not limited to linear regression and ARIMA are built and compared to select the appropriate model with highest accuracy using past stock price as training set. Besides, Twitter data has been considered to provide another angle of analysis that focus on public’s sentiment. Four Classifying models are used to value the sentiment of tweet contents separately. Extensive discussions have been presented on the performance of these models with critical evaluation.