Time Series Forecasting (Modified for higher Pytorch version & other various fixes)

Original Repository : https://github.com/laiguokun/LSTNet

Original Readme

paper

Modeling Long- and Short-Term Temporal Patterns with Deep Neural Networks.(https://arxiv.org/abs/1703.07015)

usage

You can download the datasets mentioned in the paper at https://github.com/laiguokun/multivariate-time-series-data.

We give the examples to run different datasets in ele.sh, traffic.sh, solar.sh and stoke.sh.

Environment

Python 2.7 and Pytorch 0.3.0

Docker Container Very Guided Guide for plebs

CPU VERSION

docker pull ufoym/deepo:all-py27-cpu
docker run -it ufoym/deepo:all-py27-cpu
or with vilume mounting : 
docker run -it -v Q:\git\LSTnet-demo\:/home/ ufoym/deepo:all-py27-cpu
cd /home/
git clone https://github.com/sw6-aau/LSTnet-demo.git
cd LSTnet-demo
mkdir log/ save/ data/
wget -O data/exchange_rate.txt https://sembrik.s3.eu-west-2.amazonaws.com/sw6/exchange_rate.txt
chmod +x stock-cpu.sh 
./stock-cpu.sh

GPU VERSION

docker pull ufoym/deepo:all-py27-cu90
docker run -it ufoym/deepo:all-py27-cu90
cd /home/
git clone https://github.com/sw6-aau/LSTnet-demo.git
cd LSTnet-demo
mkdir log/ save/ data/
wget -O data/exchange_rate.txt https://sembrik.s3.eu-west-2.amazonaws.com/sw6/exchange_rate.txt
chmod +x stock-gpu.sh 
./stock-gpu.sh

Example on output from training