# Generating Realistic Synthetic Population Datasets

**Authors:** Hao Wu, Yue Ning, Prithwish Chakraborty, Jilles Vreeken, Nikolaj Tatti, Naren Ramakrishnan  
**arXiv:** [1602.06844](https://arxiv.org/abs/1602.06844)  
**Date:** 2016-02-22  
**Summary page:** https://www.summarizepaper.com/en/paper/1602.06844v3/

## Key points

- The paper proposes a method to generate large datasets capturing attributes and activities of synthetic, city-level populations.
- Synthetic population datasets are essential in modern studies of societal phenomena, including epidemiology, social science, and public health.
- Synthetic population datasets are necessary to study disease propagation and intervention measures before implementation in epidemiology.
- In social science, these datasets are needed to understand how policy decisions might affect preferences and behaviors of individuals.
- In public health, synthetic population datasets capture diagnostic and procedural characteristics of patient records without violating confidentialities of individuals.
- The authors propose using the maximum entropy principle to formalize a generative model that optimally utilizes given prior information about the data while being unbiased otherwise.
- An efficient inference algorithm is designed to estimate the maximum entropy model.
- The proposed method's feasibility is demonstrated using an epidemic simulation application where synthetic population datasets are necessary for studying disease propagation and intervention measures before implementation.
- The authors' approach generates realistic synthetic population datasets that can be used for various applications in epidemiology, social science, public health, and other fields requiring large-scale data analysis.
- The proposed method's accuracy makes it a valuable contribution to the field of data analysis and modeling.

## Summary

The paper titled "Generating Realistic Synthetic Population Datasets" proposes a method to generate large datasets capturing attributes and activities of synthetic, city-level populations. Such datasets are essential in modern studies of societal phenomena, including epidemiology, social science, and public health. The authors highlight that synthetic population datasets are necessary to study disease propagation and intervention measures before implementation in epidemiology. In social science, these datasets are needed to understand how policy decisions might affect preferences and behaviors of individuals. Similarly, in public health, synthetic population datasets capture diagnostic and procedural characteristics of patient records without violating confidentialities of individuals. To generate such datasets over a large set of categorical variables, the authors propose using the maximum entropy principle to formalize a generative model that optimally utilizes given prior information about the data while being unbiased otherwise. An efficient inference algorithm is designed to estimate the maximum entropy model. The authors demonstrate how their approach is adept at estimating underlying data distributions by evaluating it against both simulated data and US census datasets. The proposed method's feasibility is further demonstrated using an epidemic simulation application where synthetic population datasets are necessary for studying disease propagation and intervention measures before implementation. The authors' approach generates realistic synthetic population datasets that can be used for various applications in epidemiology, social science, public health, and other fields requiring large-scale data analysis. In conclusion, this paper presents a novel approach to generating realistic synthetic population datasets that can be efficiently utilized for different applications across various fields. The proposed method's accuracy makes it a valuable contribution to the field of data analysis and modeling.