Traffic flow forecasting

This is a fork project from guoshnBJTU/ASTGCN-r-pytorch.

Run

Notebook: Google Colab

on PEMS04 dataset

python prepareData.py --config configurations/PEMS04_[MODEL].conf

python train_[MODEL]_r.py --config configurations/PEMS04_[MODEL].conf

We modified the ASTGCN model with full three information from last hour, yesterday and previous week.
We compare ASTGCN with GCN, LSTM, HA in various settings.

Methods	HA (h/hdw)	LSTM (h/hdw)	LSTMGC (h/hdw)	ASTGCN (h/hdw)
MAE@1h	38.75/28.15	29.46/23.91	37.1/22.52	22.53/21.84

Name		Name	Last commit message	Last commit date
Latest commit History 37 Commits
configurations		configurations
data		data
fig		fig
image		image
lib		lib
model		model
paper		paper
ASTGCN.ipynb		ASTGCN.ipynb
README.md		README.md
Test.ipynb		Test.ipynb
ha.py		ha.py
prepareData.py		prepareData.py
train_ARIMA.py		train_ARIMA.py
train_ASTGCN_full_r.py		train_ASTGCN_full_r.py
train_ASTGCN_r.py		train_ASTGCN_r.py
train_LSTMGC_full_r.py		train_LSTMGC_full_r.py
train_LSTMGC_r.py		train_LSTMGC_r.py
train_LSTM_full_r.py		train_LSTM_full_r.py
train_LSTM_r.py		train_LSTM_r.py
train_MSTGCN_r.py		train_MSTGCN_r.py
visualization.py		visualization.py