DeepST学习

DeepST学习（郑宇：深度学习在时空数据中的应用）

源代码和论文可以再这里找到

Deep Spatio-Temporal Residual Networks for Citywide Crowd Flows Prediction

https://www.microsoft.com/en-us/research/publication/deep-spatio-temporal-residual-networks-for-citywide-crowd-flows-prediction/

setup1:

python setup.py install

setup2

md5sum -c md5sum.txt

setup3:

>>> import h5py
>>> f = h5py.File('BJ16_M32x32_T30_InOut.h5')
>>> for ke in f.keys():
...     print(ke, f[ke].shape)
...
data (7220, 2, 32, 32)
date (7220,)

python -c "from deepst.datasets import stat; stat('BJ16_M32x32_T30_InOut.h5')"

=====stat=====data shape: (7220, 2, 32, 32)# of days: 162, from 2015-11-01 to 2016-04-10# of timeslots: 7776# of timeslots (available): 7220missing ratio of timeslots: 7.2%max: 1250.000, min: 0.000=====stat=====