Forest fires are threats for our ecosystems and environment because their impact is very harmful. Every year, the number of hotspots increases, indicating the increase of forest fires in some regions in Indonesia, one of them is Riau Province. To predict the hotspot occurrence, we build a web-based application based on characteristics of area using the Shiny framework. We use the C5.0 algorithm by generating tree and rule-based classification models. The Shiny framework was implemented using reactivity expression, when an input changes, the server will rebuild the output based on the input data. We use the dataset of forest fires in Rokan Hilir district, Riau Province, in 2008. The dataset consists of ten explanatory layers (physical, weather, and socio-economic characteristics) and one target layer (hotspot or non-hotspot). The implementation of the C5.0 algorithm on forest fire data resulted tree models with accuracy of 72.72% and rule-based models with accuracy of 73.51%. The output of tree models is 16 classification rules while the output of rule-based models is 15 classification rules.
