Benchmarking synthetic population quality.

To assess the viability of using synthetic populations as substitutes for real populations, we analyse the predictive performance of various machine-learning models using both real and synthetic data. Each model is tasked with predicting weight (or height) based on age, sex, and height (or weight). Case 1: the model undergoes training and testing using survey data. Case 2: an identical model is trained on synthetic population data and tested on the survey data. This table lists the accuracy of each model in predicting outcomes for linear regression and MLP regression.