LSTNet

• This repo contains an MXNet implementation of this state of the art time series forecasting model.

• You can find my blog post on the model here

Running the code

1. Download & extract the training data:

• $ mkdir data && cd data

• $ wget https://github.com/laiguokun/multivariate-time-series-data/raw/master/electricity/electricity.txt.gz

• $ gunzip electricity.txt.gz

2. Train the model (~1.5 hours on Tesla K80 GPU with default hyperparams):

• $ cd src && python lstnet.py --gpus=0

Results & Comparison

• The model in the paper predicts with h = 3 on electricity dataset, achieving RSE = 0.0906, RAE = 0.0519 and CORR = 0.9195 on test dataset

• This MXNet implementation achieves RSE = 0.0880, RAE = 0.0542 after 100 epochs on the validation dataset

• Saved model checkpoint files can be found in models/

Hyperparameters

The default arguements in lstnet.py achieve equivolent performance to the published results. For other datasets, the following hyperparameters provide a good starting point:

• q = {2^0, 2^1, ... , 2^9} (1 week is typical value)

• Convolutional num filters = {50, 100, 200}

• Convolutional kernel sizes = 6,12,18

• Recurrent state size = {50, 100, 200}

• Skip recurrent state size = {20, 50, 100}

• Skip distance = 24 (tune this based on domain knowledge)

• AR lambda = {0.1,1,10}

• Adam optimizer LR = 0.001

• Dropout after every layer = {0.1, 0.2}

• Epochs = 100