This repository contains code to download and munge historical stock data and construct neural networks to make predictions with this data
Historical data from ~3000 stocks were downloaded through Google Finance, code can be found in Download_Raw_Data.py
Momentum based performance metrics were calculated, this code can be found in Munge_Data.py
The models were constructed using Tensorflow, this code can be found in train_model.py
An individual stock price can be predicted from the saved models, this code can be found in predict_stock.py
Predictions have two forms
- Probability the stock will increase in price over a given time period
- Predicted stock price over a given time period in the future
There are four different time periods modelled:
- One day
- One week
- Two weeks
- One month