Case-Study: NLP Applications for Stock Market Prediction

Problem

Our client was a portfolio management company interested in understanding whether NLP could help them predict the stock market’s volatility.

Solution

Our approach was to contrast ideas based on the state of the art, particularly combining sentiment analysis, confidence in sentiment, intake from financial reports, and comparison versus analysts’ expectations.

As part of the work, we have delivered both a model and early prototype-level implementations of some experiments for demonstrational purposes.

Results

Albeit our main finding was that a maximum likelihood estimator is bound to agree a great deal with the analyst’s view and is not necessarily the best fit for finding outliers, which are the ones presenting promising opportunities for trading, our client found valuable insights in the NLP side of our work and proceeded to continue the development with their in-house team.