Skip to content

Wentao-Xu/IGMTF

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

10 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

IGMTF: Instance-wise Graph-based Framework for Multivariate Time Series Forecasting

The source code and data of the paper "Instance-wise Graph-based Framework for Multivariate Time Series Forecasting". image

Requirements

The framework is implemented using python3 with dependencies specified in requirements.txt.

Datasets

Overall information of datasets

Datasets Variants Timesteps Granularity Time Task Type
Traffic 862 17,544 1hour 2015 to 2016 Single-step
Electricity 321 26,304 1hour 2012 to 2014 Single-step
Exchange-Rate 8 7,588 1hour 1990 to 2016 Single-step

Reproduce the results

image

git clone https://github.com/Wentao-Xu/IGMTF.git
cd IGMTF
tar -zxvf data.tar.gz
mkdir model
  • Traffic
# Horizon 3
python learn.py --save ./model/model-traffic-3.pt --data ./data/traffic.txt --num_nodes 862 --epoch 100 --horizon 3 --hidden_size 256 --hidden_batch_size 64 --k_day 30 --n_neighbor 20

# Horizon 6
python learn.py --save ./model/model-traffic-6.pt --data ./data/traffic.txt --num_nodes 862 --epoch 100 --horizon 6 --hidden_size 256 --hidden_batch_size 64 --k_day 5 --n_neighbor 30

# Horizon 12
python learn.py --save ./model/model-traffic-12.pt --data ./data/traffic.txt --num_nodes 862 --epoch 100 --horizon 12 --hidden_size 256 --hidden_batch_size 64 --k_day 10 --n_neighbor 30

# Horizon 24
python learn.py --save ./model/model-traffic-24.pt --data ./data/traffic.txt --num_nodes 862 --epoch 100 --horizon 24 --hidden_size 256 --hidden_batch_size 64 --k_day 3 --n_neighbor 10
  • Electricity
# Horizon 3
python learn.py --save ./model/model-electricity-3.pt --data ./data/electricity.txt --num_nodes 321 --epoch 100 --horizon 3 --hidden_size 512 --k_day 5 --n_neighbor 20

# Horizon 6
python learn.py --save ./model/model-electricity-6.pt --data ./data/electricity.txt --num_nodes 321 --epoch 100 --horizon 6 --hidden_size 512 --k_day 3 --n_neighbor 3

# Horizon 12
python learn.py --save ./model/model-electricity-12.pt --data ./data/electricity.txt --num_nodes 321 --epoch 100 --horizon 12 --hidden_size 512 --k_day 10 --n_neighbor 5

# Horizon 24
python learn.py --save ./model/model-electricity-24.pt --data ./data/electricity.txt --num_nodes 321 --epoch 100 --horizon 24 --hidden_size 512 --k_day 5 --n_neighbor 20
  • Exchange-Rate
# Horizon 3
python learn.py --save ./model/model-exchange-3.pt --data ./data/exchange_rate.txt --num_nodes 8 --epoch 100 --horizon 3 --hidden_size 512 --batch_size 16 --k_day 10 --n_neighbor 10

# Horizon 6
python learn.py --save ./model/model-exchange-6.pt --data ./data/exchange_rate.txt --num_nodes 8 --epoch 100 --horizon 6 --hidden_size 512 --batch_size 16 --k_day 10 --n_neighbor 10

# Horizon 12
python learn.py --save ./model/model-exchange-12.pt --data ./data/exchange_rate.txt --num_nodes 8 --epoch 100 --horizon 12 --hidden_size 512 --batch_size 16 --k_day 10 --n_neighbor 10

# Horizon 24
python learn.py --save ./model/model-exchange-24.pt --data ./data/exchange_rate.txt --num_nodes 8 --epoch 100 --horizon 24 --hidden_size 512 --batch_size 16 --k_day 10 --n_neighbor 10

About

The source code and data of the paper "Instance-wise Graph-based Framework for Multivariate Time Series Forecasting".

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages