LSTNet simple implementation in PyTorch

A simple implementation of LSTNet and an out-of-box way about:

• Modeling
• Forecasting
• Evaluating (RMSE and MAPE)

Note:

1. Inspired by PatientEz/LSTNet_Keras. Used the same pollution dataset.
2. The original implementation of laiguokun/LSTNet used a normalization over the whole dataset, which is not standard. I normalized the training and test set separately.

Requirements:

matplotlib~=3.4.2
scikit-learn~=0.24.2
visdom~=0.1.8.9
numpy~=1.20.3
pytorch~=1.9.1

References:

• PatientEz/LSTNet_Keras
• laiguokun/LSTNet