Prediction of Dengue Fever Outbreak Based on Climate and Demographic Variables Using Extreme Gradient Boosting and Rule-Based Classification

Dengue Fever (DF) is one of the leading health problems in Indonesia. It because the number of cases tends to increase annually. This disease has also been an outbreak in almost all Indonesian regions at various periods. Prediction of DF outbreak is essential for the preparation of actions needed to prevent Extraordinary Events from occurring again in the future. This study implements the Extreme Gradient Boosting method to predict the number of dengue cases and Rule-based Classification to predict dengue fever outbreaks. Various changing scenarios were carried out, including distributing the amount of training and testing data, the type of independent variables used, the number of lag features, and hyperparameters applied to lowlands, medium lands, and highlands. Extreme Gradient Boosting produced the best dengue fever incidence prediction with a SMAPE of 27.57% in the lowlands. Meanwhile, Rule-based Classification can determine outbreak areas with the best accuracy rate of 93.75^% on medium lands. These results were accompanied by a sensitivity value of 100% and a specificity of 90.48%.