Evidence record · UDINUS · SDGs
- Unit
- Rektorat UDINUS
- Period
- 2024
- Goals
- Record
- 90815afc-b908-491a-b4e1-e9b52a94ffee
- Published
- 2026-10-30
Universitas Dian Nuswantoro conducted an activity concerning dengue fever case forecasting with intermittent data. The activity involved Anggraeni, Wiwik; Yuniarno, Eko Mulyanto; Rachmadi, Reza Fuad; Sumpeno, Surya; Pujiadi, Pujiadi; Sugiyanto, Sugiyanto; Santoso, Joan; and Purnomo, Mauridhi Hery as researchers and activity participants. The activity addressed the need for accurate forecasting of dengue cases as an early warning system to prevent future outbreaks. Forecasting dengue fever cases with intermittent data characteristics was described as rare, while obtaining good forecasting accuracy for intermittent data was identified as challenging.
The activity proposed a hybrid Empirical Mode Decomposition, Generalized Regression Neural Network, and Particle Swarm Optimization approach. Data preprocessing was first performed to ensure that the data were ready for further processing and had a relationship with dengue fever case numbers. Decomposition then extracted non-stationarity and nonlinearity patterns from each predictor variable and transformed them into several intrinsic mode functions. The intrinsic mode functions of each predictor variable were trained with a Generalized Regression Neural Network. Various data training and testing ratios and cross-validation were used to capture the best dengue fever case forecasting model. Particle Swarm Optimization was used to find optimal Generalized Regression Neural Network parameters so that the parameter-searching process became more efficient and accuracy increased.
The proposed approach was assessed for robustness and effectiveness on 21 datasets with different intermittent conditions, data periods, geographical conditions, data numbers, and data ranges. It was compared with benchmark models using mean squared error, mean absolute error, and symmetric mean absolute percentage error as evaluation indicators. The Diebold–Mariano test and pairwise sample t-test showed that the proposed model was more reliable in handling intermittent data. This activity relates to the stated goal through health-focused research on dengue fever forecasting and presents an institutional research activity addressing an identified forecasting challenge.