Predicting Localized Dengue Incidences using Ensemble System Identification

T. Loshini, Vijanth S. Asirvadam, Sarat C. Dass, Balvinder S. Gill

Research output: Chapter in Book/Report/Conference proceedingConference contribution

7 Citations (Scopus)


Neighbouring regions do have an influence on the pattern of dengue incidences for a particular site. This study presents ensemble models, where dengue incidences of a district can be estimated using a dengue prediction model of that district together with its neighbouring districts. Seven districts of Selangor are chosen in this study and an ensemble dengue incidence prediction model is built based on the respective districts and its neighbours. Dengue incidence models for each district are developed using predictor variables which include previous dengue incidences and weather variables (namely, the mean temperature, relative humidity and cumulative rainfall). These predictors are found to have specific model order lag time. To measure the efficiency of ensemble models, the formed ensemble model for each district is compared with their respective single model using the Mean Square Error (MSE) criteria. It was found that out of seven districts, five districts had better prediction accuracy based on their ensemble models. Hence, we conclude that ensemble models predict dengue incidences well.
Original languageEnglish
Title of host publication2015 International Conference on Computer, Control, Informatics and its Applications (IC3INA)
Number of pages6
ISBN (Electronic)9781479987733
Publication statusPublished - 11 Jan 2016


  • Dengue Incidences
  • System Identification
  • Mean Temperature
  • Relative Humidity
  • Rainfall
  • Mean Square Error (MSE)
  • Ensemble Model


Dive into the research topics of 'Predicting Localized Dengue Incidences using Ensemble System Identification'. Together they form a unique fingerprint.

Cite this