Stock Price Forecasting

Using Autoregressive Integrated Moving Average (ARIMA), this projects forecasts future stock prices based on the past data alone.

Getting Started

Dependencies

pip install -r requirements.txt

Training

To find the optimal values model run:

python train_agent.py --choice [ENTER A STOCK HERE]

To display a list of the available stocks run:

python train_agent.py --list show

Visualize

To visualize the data run:

python visualize_main.py --choice [ENTER STOCK HERE] --data_display [simple OR forecast]

--data_display simple

This visualizes the past stock price history, without the forecasted price

--data_display forecast

This displays the past data along with the forecasted price history of the stock

Conclusion

My model predicted the future prices for stocks in a very general sense, strongly. I believe this is not because of the model itself, but because of the market performance during and after the provided data. My model was mostly unsuccessful in ignoring the noise, even with my best efforts to station the data. This noise lead the agent to believe the market only goes up, and that is easily observed through the forecasted prices: up and to the right almost all the time.

What could be improved

As previously stated in my conclusion, the data I provided allows the agent to learn a false positive about the market; it only goes up. To improve this one could collect more data that encompasses a severe market correction.

Licence

MIT © Dylan Snyder