Pembandingan stabilitas algoritma seleksi fitur menggunakan transformasi ranking normal Vol 6, No 2, 2008

Abstract

We address the need for evaluating the ranking robustness on different classifiers in feature selection algorithm. We propose a normalized rank transformation metric to compare the stability on classifying target class on training and testing dataset. Using stability comparison is a promising effort to improve the choice decision on deploying any feature selection algorithms. We also show relationship between stability and scalability