Hybrid Model for Under-five Mortality Rate Forecasting in Nigeria
Christogonus Ifeanyichukwu Ugoh
*
Department of Statistics, Nnamdi Azikiwe University, Awka, Nigeria.
Chinwendu Alice Uzuke
Department of Statistics, Nnamdi Azikiwe University, Awka, Nigeria.
Chukwuemeka Thomas Onyia
Department of Statistics, Enugu State University of Science and Technology, Enugu, Nigeria.
*Author to whom correspondence should be addressed.
Abstract
Accurate forecasting of under-five mortality rates (U5MR) is essential for guiding public health planning and achieving child survival targets in Nigeria. This study develops optimal hybrid models for predicting U5MR using annual data from 1980 to 2022, obtained from the United Nations Inter-agency Group for Child Mortality Estimation (UN IGME). Five individual models—autoregressive integrated moving average (ARIMA), exponential smoothing state space (ETS), multilayer perceptron (MLP), Prophet, and extreme gradient boosting (XGBoost)—were assessed alongside four hybrid models: ARIMA-ETS, ARIMA-MLP, ARIMA-Prophet, and ARIMA-XGBoost. Hybrid forecasts were generated using a weighted averaging method, with model weights optimized through a genetic algorithm. Model performance was evaluated using RMSE and MAPE on an out-of-sample test set (2018–2022), with sensitivity analysis performed over three, five, and seven-year forecast horizons. Results indicated that hybrid models, particularly ARIMA-ETS and ARIMA-Prophet, significantly outperformed individual models across all metrics and timeframes, for both male and female U5MR. This study recommends that national agencies adopt these hybrid models for mortality forecasting and planning.
Keywords: Under-five mortality rate, hybrid forecasting models, time series analysis, genetic algorithm, Nigeria