Analyzing Sentiment of Stock Market Data from Twitter Using Machine Learning and Deep Learning Techniques

A widely accepted theory in stock prediction called Efficient Market Hypothesis (EMH) has failed in modern times, and since then the research focus had took a sharp turn toward machine learning, deep learning and other statistical techniques. This paper is about taking non-quantifiable data such as tweets on a company’s stocks and the tweets about the stock market that mention the company in them and predicting its future stock trend with Twitter sentiment classification. We carry out this study by assuming tweets have/show impact on the behavior of stock market. To show this, we will create a classification model which depict polarity of tweets being positive or negative and we will compare the findings of the same with the existing approaches. We will compare the existing models Support Vector Machine (SVM) and LSTM. The text is preprocessed into the relevant form to feed the algorithms, this is done by Natural Language Processing (NLP) techniques. Linear methods like machine learning and nonlinear methods like deep learning models are being compared. For our dataset, the ML model (SVM) performed better than LSTM (Deep Learning Approach).