This package (GP-DLM) is the group pooling deep learning algorithm for tourism demand forecasting. Please be aware that:

• The pooling stage will need run the extra dynamic time warping clustering for generating the pooling group.
• Pooling data is generated separately.
• The training of DLM needs extra efforts based on specific data set, and direct running of the provided code DOES NOT always generate the promised performance.
• For the training of the model on the data set, please spend your own patient time and the code publisher will NOT provide assistance on this issue.

If you use it for a scientific publication, please include a reference to this paper.

• Yishuo Zhang, Gang Li, Birgit Muskat, Rob Law and Yating Yang(2020). Group Pooling For Deep Tourism Demand Forecasting. Annals of Tourism Research, Vol 82, May 2020

BibTex information:

@article{ZLMLY2020,
title = {Group Pooling For Deep Tourism Demand Forecasting},
volume = {82},
doi = {10.1016/j.annals.2020.102899},
journal = {Annals of Tourism Research},
author = {Zhang, Yishuo and Li, Gang and Muskat, Birgit and Law, Rob and Yang, Yating},
month = May,
year = {2020},
keywords = {tourism demand forecasting, AI-based methodology, group-pooling method, deep-learning model, tourism demand similarity, Asia Pacific travel patterns}, 
}

The related dataset HK-MO2018 for above paper can be found at TULIP Lab Open-Data:

• HK-MO2018

• Python 3.6
• Keras 2.2 (2.3 won't work for the weights saving)
• Tensorflow

• Setting up the pooling data by using Dynamic_time_warping.py and pooling.py
• Feeding the pooled data into DLM by using DLM.py and configuration.py
• Run the forecasting.py