Stock price prediction with RNN. The data we used is from the Chinese stock.
- Python 3.5
- TuShare 0.7.4
- Pandas 0.19.2
- Keras 1.2.2
- Numpy 1.12.0
- scikit-learn 0.18.1
- TensorFlow 1.0 (GPU version recommended)
I personally recommend you to use Anaconda to build your virtual environment. And the program probably cost a significant time if you are not using the GPU version Tensorflow.
You can run fetch_data.py to get a piece of test data. Without changing the script, you can get two seperated csv file named:
000002-from-1995-01-01.csv=====> Contains general data for stock 000002 from 1995-01-01 to today.000002-3-year.csv=====> Contains candlestick chart data for stock 000002 (万科A) for the most recent 3 years.
You are expected to see results look like (the first DataFrame contains general data where the the second contains detailed candlestick chart data):
$ python3 fetch_data.py
[Getting data:]#########################################################################################
Saving DataFrame:
open high low volume amount close
0 20.64 20.64 20.37 16362363.0 3.350027e+08 20.56
1 20.92 20.92 20.60 21850597.0 4.520071e+08 20.64
2 21.00 21.15 20.72 26910139.0 5.628396e+08 20.94
3 20.70 21.57 20.70 64585536.0 1.363421e+09 21.02
4 20.60 20.70 20.20 45886018.0 9.382043e+08 20.70
Saving DataFrame:
open high low volume price_change p_change ma5 ma10 \
0 20.64 20.64 20.37 163623.62 -0.08 -0.39 20.772 20.721
1 20.92 20.92 20.60 218505.95 -0.30 -1.43 20.780 20.718
2 21.00 21.15 20.72 269101.41 -0.08 -0.38 20.812 20.755
3 20.70 21.57 20.70 645855.38 0.32 1.55 20.782 20.788
4 20.60 20.70 20.20 458860.16 0.10 0.48 20.694 20.806
ma20 v_ma5 v_ma10 v_ma20 close
0 20.954 351189.30 388345.91 394078.37 20.56
1 20.990 373384.46 403747.59 411728.38 20.64
2 21.022 392464.55 405000.55 426124.42 20.94
3 21.054 445386.85 403945.59 473166.37 21.02
4 21.038 486615.13 378825.52 461835.35 20.70
