How to Efficiently Predict Dengue Incidence in Kuala Lumpur

Abstract

Mosquito-borne diseases are rapidly spreading in all regions of the world with an estimation of 2.5 billion people globally are at risk. The recent surge in dengue outbreaks has caused severe affliction to Malaysian society. Hence, the ability to predict a dengue outbreak and mitigate its damage and loss proactively is very critical. In this paper, we study the possibility of applying machine learning (ML) and deep learning (DL) approaches to predict the number of confirmed dengue fever (DF) cases in Kuala Lumpur. We identified several contribution factors correlate to a dengue outbreak. In addition to the two frequently used factors (daily mean temperature and daily rainfall), we also took into account the enhanced vegetation index (EVI), humidity and wind speed as input factors to our prediction engines. We collected and cleansed data on these factors and the daily DF incidents in Kuala Lumpur from 2002 to 2012. We then used these data to train and evaluate our 3 ML/DL models. Among the three models, GA_RNN was the best performer and achieved a MAE of 10.95 for DF incidence prediction.