Deep Learning-Based Sentiment Analysis for Predicting Financial Movements

Abstract

Sentiment analysis is a computational study of opinions, feelings, emotions, ratings and attitudes towards entities such as products, services, organizations, individuals, issues, events, subjects and their attributes. Our research is used to predict stock market movements, aims to improve the accuracy of polarity of comments, in order to accurately predict financial movements. This by creating a dictionary of emojis that contains the emoji as keys and its meanings as values. We used the dictionary at the preprocessing level to keep the meanings of emojis because they carry a lot of emotions that help us to clearly specify the polarity of the comments. We have also created a list of stopwords related to the financial field to properly clean our database. Time series and linking sequences of data is very important to properly predict stock market movements. We have therefore chosen to work with the Long short-term memory (LSTM) model. Next, we came up with two models: the first model to predict stock market movements using investor sentiment analysis of Amazon stock which gives us 93% accuracy. The second model is used to predict financial movements through historical Amazon prices. We extracted the database we used for the sentiment analysis from Twitter as the Twitter comments are up to date. As for the historical prices of Amazon stock we extracted from the most famous trading platform YahooFinance.