# Generating Synthetic Population

**Authors:** Bhavesh Neekhra, Kshitij Kapoor, Debayan Gupta  
**arXiv:** [2209.09961](https://arxiv.org/abs/2209.09961)  
**Date:** 2022-09-20  
**Summary page:** https://www.summarizepaper.com/en/arxiv-id/2209.09961v1/

## Key points

- The paper presents a method for generating synthetic populations at various administrative levels for India
- Machine learning and statistical methods are used to create the synthetic population from survey data such as Census of India 2011, IHDS-II, NSS-68th round, GPW etc.
- Characteristics of individuals in the population include age, gender, height, weight, home and work location, household structure, preexisting health conditions, socio-economical status and employment
- Schools are assigned to children based on a probability-based approach that considers the distance between an individual's home and available schools
- Workplaces are assigned to adults based on their job label by randomly selecting from a suitable subset of allowed workplaces
- Additional attributes such as public transport usage and essential worker status are included in the synthetic population
- The authors generate the synthetic population for various districts of India and compare it with source data using metrics such as Bhattacharya distance and statistical likelihood techniques (CS-test and KS-test)
- Experiment results show that the generated synthetic data can realistically simulate the population for various districts of India
- This study has significant implications for policy makers who require accurate demographic information at different administrative levels but may not have access to complete or up-to-date survey data.

## Summary

This paper presents a method for generating synthetic populations at various administrative levels for a country like India. The synthetic population is created using machine learning and statistical methods applied to survey data such as Census of India 2011, IHDS-II, NSS-68th round, GPW etc. The generated synthetic population defines individuals in the population with characteristics such as age, gender, height, weight, home and work location, household structure, preexisting health conditions, socio-economical status and employment. To assign schools to children in the synthetic population the authors use a probability-based approach that takes into account the distance between an individual's home and available schools. Similarly workplaces are assigned to adults based on their job label by randomly selecting from a suitable subset of allowed workplaces. Additional attributes such as whether an individual uses public transport or is an essential worker are also included in the synthetic population. The authors generate the synthetic population for various districts of India and compare it with source data using metrics such as Bhattacharya distance to quantify similarity of joint age-height and age-weight distributions; statistical likelihood techniques (CS-test and KS-test); visual comparison of geographical spread of households, schools and workplaces; among others. The experiment results show that the generated synthetic data can realistically simulate the population for various districts of India. This study has significant implications for policy makers who require accurate demographic information at different administrative levels but may not have access to complete or up-to-date survey data. The proposed method provides a reliable alternative that can be used to generate realistic demographic information required for planning purposes.