AAPL predictions

Inspiration

Two CS and Econ students that were very excited about the Reddit IPO, just to find out it wasn't available for non-Americans

What it does

Input a stock on the S&P 500, and watch stockr.ai predict its close over the next 30 days.

How we built it

Stock Data pulled from the Yahoo! Finance API, Used tensorflow to build our GRU model, used matplot for the graphing. Used Flask, HTML + CSS for the webapp.

Challenges we ran into

Training. It's very hard to accurately predict stock prices on close price alone.

Accomplishments that we're proud of

We could attain < 1.5 RMSE on certain stocks during the training and validation (AAPL, GIS, MMM). Additionally, the future stock price predictions were (arguably) accurate in shape for stocks that had low volatility (a good start nonetheless).

What we learned

Machine Learning is all about trial and error. Changing the models, changing the epoch, changing the data and the hyperparameters.

What's next for stockr.ai - stock predictor

Increasing the model flexibility - have more layers or more units in the existing layers. Add more features to improve accuracy - in the future consider opening price, highs, lows, volume etc (Far future) Be able to pull trends from news headlines to more accurately predict.