Data Driven Prediction of Dengue Incidence in Thailand

摘要

Communicable diseases such as dengue pose a significant threat on public health across the world. Modeling an accurate and efficient prediction of dengue disease will improve public health response planning to outbreaks. However, despite the fact that many researches has focused on dengue prediction, it has been lacking geographical variation of dengue fever taken into account. Dengue is a mosquito-borne virus that annually infects over 400 million people worldwide. The infection pattern is different from region to region. We developed a model for predicting dengue fever for four provinces of Thailand with geographical variation taken into account. These predictions show slightly varying outcomes across provinces. Support Vector Regression (SVR) was used as the modeling tool. Additionally, we introduced a novel method of assessing regression model in terms of accuracies over Mean Square Error (MSE) which does not capture the behavior of data pattern spatially. This novel method resulted in 71% accuracy of prediction for Kamphaeng Phet province. The proposed model of prediction facilitates administrative bodies to make informed decisions in the context of public health of Thailand.