Modelling children ever born using performance evaluation metrics: A dataset

Abstract Predicting the number of total children ever born in a country is a key component for proper implementation of economic growth policy. Here, performance metrics were used to predict models that appropriately describe the factors that affect children ever born. A comparison of 60% training and 40% validation, 70% training and 30% validation, 80% training and 20% validation also 90% training and 10% validation was performed respectively to examine the three models’ behaviours (Poisson regression, Negative Binomial regression and Generalized Poisson regression) with RMSE, R2, MAE and MSE as performance metrics. Although all the three models had almost identical performance evaluation metrics, the Poisson regression was chosen as the most appropriate model because it is the simplest model.