In this project, we aim to enhance the prediction of stock market movements using sentiment analysis and deep learning. We divide this project into four phases. In the first part, we aim to find textual data in tweets, comments, etc. In the second phase, pre-trained language models are used to generate sentence-level embeddings for each sample in the dataset. Next, in the third section, a classifier is trained on the embeddings and predicts sentiments for all text data. In addition, each predicted sentiment is a single number indicating how positive the collective data has been for that stock on that particular day. Finally, in the last part, we extract price fluctuations for each stock symbol on each day and generate technical features. Moreover, sentiment predictions are also added to the technical features; we then find the best features and train various hybrid deep learning models to predict the stock price movement for the next day. In this project, we have tested our approaches to a total of 24 Nasdaq stocks.

The results and methodology are available in the report section.

Here is a youtube link for this project: https://youtu.be/D6BLZUh3QHY

This Projct is developed by: Sepehr Asgarian and Rouzbeh MeshkinNejad