Application of Autoregressive Integrated Moving Average Model and Weighted Markov Chains on Forecasting Under-Five Mortality Rates in Nigeria

Abstract

The aim of this paper is to obtain the best model that will be used to predict Under-Five Mortality Rate (U5MR) between Autoregressive Integrated Moving Average (ARIMA) model and Weighted Markov Chains (WMC). The annual dataset of U5MR in Nigeria for the period 1980-2019 is obtained from the official website of World Bank. The descriptive statistics and the unit root test for the stationarity of data were carried on the data series. ARIMA was modelled to U5MR using the techniques of Box-Jenkins while WMC was modelled using the techniques of k-means cluster analysis, Chi-Square, and Correlation. The best ARIMA model was obtained using Bayesian Information Criterion (BIC) while the best forecast model was obtained using Theil’s U Statistics and Mean Absolute Percentage Error (MAPE). U5MR attained stationarity after third differencing under ARIMA model dynamics. ARIMA(0,3,2) is considered the best ARIMA model with BIC of -2.679, and was selected as the best forecast model with Theil’s U Statistic of 0.000014 and MAPE of 0.174336%. The fitted model was used to make out-sample forecast for the period 2020-2030, which showed a steady decline. The findings of this paper will help in establishment and implementation health policies.