Developing Spatial Decision Tree using NBTree Algorithm for Hotspot Distribution in Riau.

Abstract

One of environment problems that emerged almost every year in Indonesia is forest fire in many regions including Riau. Forest fires cause negative impacts to the environment, such as transboundary haze pollution, health and life problem, and material damages. This phenomenon effects not only on community in Riau but also in neighbor coutries such as Singapore and Malay. Hotspot distribution is important for preventing occurrence of forest and land fire. Hotspot distribution data can be analyzed using spatial data mining techniques. One of techniques in spatial data mining is spatial decision tree. Spatial decision tree will build classifier from spatial data that can be used to create classification rules. This research aimed to develop classifier to predict the amount of hotspot in a territory according to several thematic layers: industrial timber plantation, logging concession, and land use. This research uses spatial join index and complete operator to provide a spatial dataset in which the conventional spatial decision tree algorithm can be applied to the data set. Then, NBTree algorithm will be used to develop a spatial decision tree. The result of this research is a classifier that contains 17 classification rules with accuracy 68,35%. Keywords: spatial decision tree, spatial join index, complete operator, NBTree algorithm