This repository contains code written for the master thesis titled: "Predicting the stock market with sentiment analysis of newspaper text".
Author: Juan Luis Ruiz-Tagle
Director: Emilio Serrano
- I am researching on the capabilities that NLP models have on predicting stock market returns.
- I have chosen Tesla and Bitcoin (not a stock, but does the job), since they are highly volatile and prone to change due to sudden hypes
- I use BERT, a general purpose NLP network developed by google, to predict the sentiment of news articles related to these stocks and trying to find correlations with their returns
Next steps are using conditional Generative Adversarial Networks (cGANs) to combine BERT's sentiment analysis with quantitative techniques to give more accurate predictions.
Here you can see the model price predictions for Bitcoin and Tesla based solely on newspaper text