Studi Perbandingan Penggabungan Metode Pemilihan Fitur dengan Metode Klasifikasi dalam Klasifikasi Teks

Abstract

The main purposes of this comparative study is to obtain the best features and the method of selecting the most suitable for a particular classification method, as well as provides an overview of the performance and the accuracy of each selection features when combined with any method of classification.  From the experiment it shows that for Naive Bayes classification method has the maximum degree of accuracy when combined with feature selection using Support Vector Machine. K-Nearest Neighbor classification obtains maximum accuracy when it is combined with feature selection using Information Gain and Uncertainty, with the value of k is 4. Furthermore, for Neural Network classifier, it looks less when it is combined with the feature selection tested since it is only produce maximum accuracy less than 50% combined with Information Gain. Moreover, Support Vector Machine resulting maximum accuracy when it is tested using Information Gain, Chi Squared, Deviation and SVM.