PyTre supports the development and testing of Travel Behavior embeddings. In its first version, PyTre can be used to encode categorical variables.
If you’re interested in using PyTre, you’ll need a few Python packages installed:
- General stuff: pandas, sklearn, numpy, matplotlib, pickle, scipy, tempfile, adjustText
- Deep Learning: tensorflow, keras
- For using the notebooks: collections, pylogit
To understand and try the concept of travel embeddings, we recommend to read the paper "Rethinking travel behavior modeling representations through embeddings" (https://arxiv.org/abs/1909.00154), and to run the "PyTre_swissmetro_paper.ipynb” notebook.
If you want to use this in your research, please cite:
"F. C. Pereira, Rethinking travel behavior modeling representations through embeddings, arXiv e-prints (2019) arXiv:1909.00154arXiv:1909.00154."