A spatial decision tree based on topological relationships for classifying hotspot occurences in Bengkalis Riau Indonesia

Abstract

Forest fires in Riau province Indonesia, are frequently occurred every year especially in dry seasons. Hotspot is an indicator for forest fire events. Hotspots monitoring is an activity to prevent forest fires. Hotspot data are spatial data that are represented in points. In order to analyze the data, spatial algorithms are required. The extended spatial ID3 algorithm is a spatial classification algorithm for creating a spatial decision tree from spatial datasets. This research applied the extended spatial ID3 algorithm on the forest fires data in Bengkalis district, Riau province Indonesia. The data include hotspots and non-hotspots, weather data, socio-economic data, and geographical characteristics of the study area. The result of this research is a decision tree with the income source layer as the label of root node. As many 137 classification rules were generated from the tree. The accuracy of the tree is 75.66% on the forest fires dataset in Bengkalis district, Riau province.