Twitter Sentiment Trading

The team for this project explored the use of sentiment analysis on financial tweets on Twitter. Sentiment Analsyis is a branch of Natural Language Processing that involves determining the sentiment of text - in this case whether a tweet is positive or negative (bullish or bearish) on financial twitter data. In doing so, they explored the use of different classification methods (SVMs, Naive Bayes) and word embedding methods such as FinBert and ULMFiT (since words are text, and machine learning models require numeric data, text needs to be converted into a suitable numeric representation - word embeddings). Finally, they backtested a simple trading strategy based on this derived sentiment.

A Google Colab notebook is available here

And the slides to their presentation are available here

Github Repo