Stock price movements are said to be unpredictable. We want to test out if machine learning and big data can do what only a few well-trained financial analysts can: predict stock price movements based on recent macroeconomic news.
What it does
Scraper module takes text content from news pages. The data is sent to sentiment analysis API to compute a sentiment score that tells if the news article is about positive or negative news. Then we aggregate these data to give one global score to determine buy or sell.
How we built it
Used google and x-ray node modules to build a scraper module. Sentiment Analysis API to compute the sentiment score. node.js for server setup.
Challenges we ran into
Implementation and incorporation of machine learning into finance: the joint-knowledge and new paradigm were required from all teammembers.
Accomplishments that we are proud of
We were able to work together even if we all come from different schools and it was the first time we've met each other. Also, despite of having two first-time hackers, we could all work together to build a fully-functional application.
What we learned
What's next for Stock It
Compute theoretical fair price of the stock using Capital Asset Pricing Model (and other valuation models) and compare it to current value of the stock to determine the theoretical mispricing and potentially the machine learning algorithm will be able to tell what is the reason behind that mispricing. Furthermore, this application could be used for predicting option prices, currencies, and even political event such as elections.