niuyishuai/STNN_Disease

Predict the Distribution of Disease by RNN and STNN

This project is supported by the National Natural Science Foundation of China (Grant No: 11601327) and the Key Construction National “985” Program of China (Grant No: WF220426001).

Data

Aids

The file aids.csv contains the raw temperature data. The 156 rows correspond to the 156 timestep, and the 29 columns are the 29 space points. The file aids_relations.csv contains the spatial relation between the 29 space points. It is a 29 by 29 adjacency matrix A, where A(i, j) = 1 means that series i is a direct neighbor of series j in space, and is 0 otherwise.

Flu

The file flu.csv contains the raw temperature data. The 156 rows correspond to the 156 timestep, and the 29 columns are the 29 space points. The file flu_relations.csv contains the spatial relation between the 29 space points. It is a 29 by 29 adjacency matrix A, where A(i, j) = 1 means that series i is a direct neighbor of series j in space, and is 0 otherwise.

Heat

The file heat.csv contains the raw temperature data. The 200 rows correspond to the 200 timestep, and the 41 columns are the 41 space points. The file heat_relations.csv contains the spatial relation between the 41 space points. It is a 41 by 41 adjacency matrix A, where A(i, j) = 1 means that series i is a direct neighbor of series j in space, and is 0 otherwise.

Model

Spatio-Temporal Neural Networks for Space-Time Series Forecasting and Relation Discovery

ICDM 2018 - IEEE International Conference on Data Mining series (ICDM)

Conference Paper

Journal Extension

Commands for reproducing synthetic experiments:

STNN

python train_stnn.py --dataset aids --outputdir output_aids --manualSeed 1932 --xp stnn

python train_stnn.py --dataset flu --outputdir output_flu --manualSeed 7011 --xp stnn

python train_stnn.py --dataset heat --outputdir output_heat --manualSeed 2021 --xp stnn

STNN-R(efine)

python train_stnn.py --dataset aids --outputdir output_aids --manualSeed 3301 --xp stnn_r --mode refine --patience 800 --l1_rel 1e-8

python train_stnn.py --dataset flu --outputdir output_flu --manualSeed 3796 --xp stnn_r --mode refine --patience 800 --l1_rel 1e-8

python train_stnn.py --dataset heat --outputdir output_heat --manualSeed 5718 --xp stnn_r --mode refine --patience 800 --l1_rel 1e-8

STNN-D(iscovery)

python train_stnn.py --dataset aids --outputdir output_aids --manualSeed 1290 --xp stnn_d --mode discover --patience 1000 --l1_rel 3e-6

python train_stnn.py --dataset flu --outputdir output_flu --manualSeed 8837 --xp stnn_d --mode discover --patience 1000 --l1_rel 3e-6

python train_stnn.py --dataset heat --outputdir output_heat --manualSeed 9690 --xp stnn_d --mode discover --patience 1000 --l1_rel 3e-6

RNN

Here LSTM and GRU are used.

Commands for reproducing synthetic experiments:

LSTM

python train_rnn.py --dataset aids --model LSTM --manualSeed 1208 --xp LSTM_aids

python train_rnn.py --dataset flu --model LSTM --manualSeed 1471 --xp LSTM_flu

python train_rnn.py --dataset heat --model LSTM --manualSeed 6131 --xp LSTM_heat

GRU

python train_rnn.py --dataset aids --model GRU --manualSeed 1208 --xp GRU_aids

python train_rnn.py --dataset flu --model GRU --manualSeed 1471 --xp GRU_flu

python train_rnn.py --dataset heat --model GRU --manualSeed 6131 --xp GRU_heat