Comparative Modeling of Measles Incidences Among Children Below Five Years in Rivers State, Nigeria Using the Holt-Winters, SARIMA, and Hybrid SARIMA-ANN Models

This research analyzed various time-series forecasting models for measles occurrences in children younger than five in Rivers State, Nigeria, utilizing monthly data sourced from the Rivers State Hospital Management Board and allied partners spanning from January 2015 to December 2024. Python 3.11 software was used for the data analysis. The investigation focused on the temporal patterns of measles cases and assessed the prediction efficacy of additive Holt–Winters, multiplicative Holt–Winters, Seasonal Autoregressive Integrated Moving Average (SARIMA), and hybrid SARIMA–Artificial Neural Network (ANN) models. To explore the data series, methods such as descriptive statistics, seasonal decomposition, the Jarque–Bera normality examination, the Augmented Dickey–Fuller unit-root evaluation, autocorrelation assessment, and the Ljung–Box test were utilized. The forecasting efficiency was measured via root mean squared error (RMSE), mean absolute scaled error (MASE), and mean absolute percentage error (MAPE). Findings indicated an average of 26.17 monthly cases of measles, with a standard deviation of 21.43 and a peak of 124 cases. The data series displayed positive skewness and did not conform to a normal distribution, while the results from the Augmented Dickey–Fuller test suggested stationarity at level significance (p < 0.001). The Ljung–Box analysis uncovered significant serial correlations at lags 12 and 20 (p < 0.001). Out of the four forecasting techniques, the multiplicative Holt–Winters method yielded the best results, achieving the lowest RMSE at 16.43 and a reported MASE of 86.68. Conversely, the hybrid SARIMA–ANN approach was the least effective, presenting an RMSE of 22.05 and a reported MASE of 261.66. The study concludes that the multiplicative Holt–Winters method stands out as the most effective model among those tested. It suggests the validation of the reported MASE values, the importance of consistent out-of-sample evaluation, and enhancing regular measles monitoring to aid in timely forecasting, vaccination strategy, and focused public health actions.