RRM-PoE

An example of a product of experts (PoE) choice model using BIOGEME

Product of experts uses the product function to generate the utility function:

\frac{1}{Z} \prod p(y_j|y_k)f(y_k)

Random regret models can be formulated as a PoE which is shown in this project

File description

Model files:

• rrm-poe.py Product of Experts model to calculate a random regret model
• rrm-poe2.py alternative variant of the PoE model
• rrm2010.py original formulation of the regret function (RRM2010 paper)
• logit.py standard MNL model
• mixedlogit.py mixed logit model

HTML files: model estimation results

Dataset

biogeme_trh.csv

Dataset tech report: Sp_TrainHotel_Draft 5_Oct 27.2016.pdf

Getting started

To run, use python3 rrm-poe.py. Optionally, to test the dataset on different models, replace rrm-poe.py with the appropriate python script.

rrm-poe2.py uses a different mathematical function to calculate the product utility function. Theoretically it should result in the same expression and estimated parameters. However, due to the internal optimization calculation, it may or may not produce the same result as rrm-poe.py.

Prerequisites

Python 3.5+ (with pip3), PythonBiogeme Version 2.5 download

Note: PandasBiogeme is not supported yet

Versioning

0.1 Inital version

Authors

Melvin Wong (Github)

Licence

See LICENCE for details