Adaptive Synthetic Population Generation using Gibbs Sampler

This repository provides a reproducible implementation of the adaptive synthetic population generation framework described in the paper "Adaptive synthetic generation using Gibbs Sampler".

The method combines:

• Initial population generation using a Gibbs Sampler
• Annual demographic projection through dynamic projection
• Adaptive resampling using Gibbs-based correction based on newly available data

---

Repository Overview

This repository includes three Jupyter notebooks:

• 01_generation.ipynb: Generates an initial synthetic population based on disaggregated data using a Gibbs sampler.
• 02_projection.ipynb: Simulates year-by-year demographic evolution using event-driven microsimulation.
• 03_resampling.ipynb: Applies a resampling step to align synthetic data with newly available marginal distributions using Gibbs resampling.

---

Note: The original Swiss MTMC datasets used in the paper are not included due to data sharing agreements. A dummy dataset is provided for demonstration purposes.

---

1. Install dependencies

You can install required Python packages using: pip install -r requirements.txt

2. Run the Notebooks

Use Jupyter Notebook to execute each notebook in order:

1. 01_generation.ipynb
2. 02_projection.ipynb
3. 03_resampling.ipynb

Each notebook is self-contained and includes documentation to help reproduce the steps from the paper.

---

Demo Data

We include a small dummy dataset: synthetic_sample.csv to illustrate the workflow.
This dataset mimics the structure of the real MTMC data and supports full execution of the code.

Attributes used

Individual level:

• Age (grouped)
• Sex
• Employment status
• Marital status
• Driving license

Household level:

• Household size
• Household type
• Number of cars

You may replace the dummy data with your own disaggregated dataset, as long as the structure and attributes are preserved.

---

Citation

Paper: Adaptive synthetic generation using one-step Gibbs Sampler
Authors: Marija Kukic, Michel Bierlaire
Status: Published in Transportation Research Interdisciplinary Perspectives
Link: https://doi.org/10.1016/j.trip.2025.101597

If you use this code for academic work, please cite the paper.

---

Contact

For any additional questions, you can email me! Marija Kukic
marija.kukic@epfl.ch