EVALUATION OF CLUSTERING SEASON ZONE IN NGAWI EAST JAVA USING SUPPORT VECTOR MACHINE (SVM)
Bambang Widjanarko Otok, Agus Suharsono, Elly Nur Shobibah · 2011
Ngawi is the fifth-largest district that contribute 5, 7 4% of the total rice production in East Java Indonesia. Thus, it is needed to find information about weather prediction. First steps to increase the accuration of weather predictions is done by accurately classifying Season Zone (ZOM). Until! now, BMKG [I} still use the complete linkage method to create a grouping of ZOM. While one method of classification which is currently growing rapidly is the non parametric method of Support Vector Machine (SVM) [2}. SVM method One Againts One Strategy is proposed to evaluate the classification accuracy ofZOM, because they have very good perfomance in the process of classification for multiclass SVM. The result of clustering based on rainfall data that has been reduced with factor analysis produced 4 ZOM. The result of clustering evaluation using SVM method provides information that classification accuracy of ZOM after having modification by elevation map is equal to 50%. Meanwhile, the classification accuracy of BMKG 's ZOM is equal to 72.2222%. The classification accuracy of ZOM after having modification is lower because clustering results still tend to be subjective, so there is not guarantee to the homogenity.