Application of Group Method of Data Handling Methodology for the Future Prediction of Birth Rate: A Case Study of India

Abstract

The present study aimed at comparing the group method of data handling type artificial neural network (GMDH-ANN) model with several time series models, namely the moving average (MA), double exponential smoothing (DES), autoregressive (AR) and autoregressive integrated moving average (ARIMA) for the future prediction of India’s birth rate. In this study, time series data concerning the birth rate was collected from the period 1995–96 to 2019–2020. The coefficient of determination (R²), mean absolute error (MAE), mean absolute percentage error (MAPE), mean squared error (MSE) and root mean square error (RMSE) have been used to compare the performance of various considered models. Based on these criteria’s GMDH-type ANN model performs better than other conventional statistical models. Thus, the GMDH-type ANN model was used for the future prediction of the birth rate over the next 20 years. The work done in this study will be immensely useful for the government in allocating resources and planning the future for children's services that are based on the expected fertility rate.