Stock trading is a lucrative, but very risky and unpredictable, way to make money. If we can remove or reduce some of that uncertainty it could encourage more people to get involved in the market.
What it does
Deep Money takes the input of what the price of the stock has been for the past and makes predictions on what the price will be in the future. The prediction can be for what the stock price will be over the next few days or next few minutes depending on what the trader wants.
How we built it
Firstly, we retrieve past stock data from the Alpha Vantage API. We then "clean" that data to make it into training data for our neural network. After training the neural network, we have bots, simulating investors, with a set "wallet", follow the neural networks recommendations on a certain stock to see how much money the bot makes or loses.
Challenges we ran into
Getting the stock data into a form for training our neural networks was a more challenging task than we had originally anticipated. Setting up PyTorch was also a non-trivial task as well.
Accomplishments that we're proud of
We did get the data into useable form and wrote scripts that could get us stock data for any index, process that data, and generate a usable csv file for training. We also got the Neural Network to roughly 40% accuracy.
What we learned
We learned how to process stock data into the form we wanted it in using Python and a few Python modules, such as csv and pandas. We also learned more about PyTorch and Deep Learning Studio.
What's next for Deep Money
Improve the neural network, add some more data inputs outside of past stock values, and streamline the visualization and testing process.