Predict if the next 30 days mean close price of a certain stock will be higher/lower than it was during the past 30 days.
algorithm details (click to expand)
- gather S&P500 OHLCV data and split into train/test
sp500forecaster [DEBUG]: collected 404 train, 101 test sp500 stocks
sp500forecaster [DEBUG]: building train time windows- transform normalized OHLCV train data into visual time windows
- train CNN forecaster
- evaluate on test time windows
usage: create.py [-h] [-o OUTPUT] [-v] [-d] stocknum epochs
Create a model to predict future stock trends.
positional arguments:
stocknum number of sp500 stocks to retrieve (0=all)
epochs number of training epochs
optional arguments:
-h, --help show this help message and exit
-o OUTPUT, path to output directory (default ./out)
-v, --verbose verbose output, default true
-d, --debug visual train info, default falsePretrained weights:
| weights | backtest_oa | stocknum |
epochs |
samples |
|---|---|---|---|---|
| n0_acc0.75_20180918231542.h5 | 0.79 | 0 (all) | 10 | ~259k train, ~65k val, ~85k test |
usage: predict.py [-h] [-o OUTPUT] [-v] weights symbols [symbols ...]
Predict future stock trends using a pretrained forecaster.
positional arguments:
weights path to h5 forecaster weights
symbols list of 1 or more iex-supported tickers
optional arguments:
-h, --help show this help message and exit
-o OUTPUT, --output OUTPUT
path to output directory (default ./out)
-v, --verbose verbose output, default true
Example:
$ python predict.py res/n0_acc0.75_20180918231542.h5 AAPL MMM
Using TensorFlow backend.
sp500forecaster [DEBUG]: symbols: ['AAPL', 'MMM']
sp500forecaster [DEBUG]: processing AAPL
sp500forecaster [DEBUG]: positive future prediction for symbol AAPL
sp500forecaster [DEBUG]: processing MMM
sp500forecaster [DEBUG]: positive future prediction for symbol MMM$ cat /etc/issue*
Ubuntu 16.04.4 LTS
$ python --version
Python 2.7.12
$ pip install -r requirements.txt
...