LEAST SQUARES SUPPORT VECTOR MACHINE UNTUK KLASIFIKASI DATA MULTIKELAS
Nurkamila Jafar · Hasanuddin University Repository · 2016
Least Squares Support Vector Machine (LS-SVM) is one variation of Support Vector Machine (SVM) to clasify the large data in data mining. SVM solve the problem of programming quadratic, while the LS-SVM solve linear equations. Initially SVM only used for clasify binary, then develop some methods for clasify multiclass data. This research used One Against All method and One Against One method with RBF kernel, polynomial kernel, and linear kernel. Both methods will compared to see the accuracy each its kernel, as well as the sum of missclassification using confusion matrix. Experiment assessing the accuracy of both methods on two dataset. The research results showed that the method of One Against One is better than the One Against All. These results are show the accuracy for each kernel RBF is 100%, polinomial is 100%, and linear is 95,556% use One Against All method for dataset 1.Kernel RBF is 26,531%, polinomial is 20,635%, and linear is 15,193% use One Against All method for dataset 2. While, the accuracy for One Against One method of each kernel, kernel RBF, polinomial, and linear are 100% for dataset 1. Kernel RBF is 47,166%, polinomial is 50,794%, and linear is 44,218% for dataset 2.