Optimalisasi Kinerja Klasifikasi Melalui Seleksi Fitur dan AdaBoost dalam Penanganan Ketidakseimbangan Kelas

Abstract

One of the problems in data mining classification is class imbalance, where the number of instances in the majority class is more than the minority class. In the classification process, minority classes are often misclassified, because machine learning prioritizes the majority class and ignores the minority class so that this can cause the classification performance to be not optimal. The purpose of this study is to provide a solution to overcome class imbalances so as to optimize classification performance using chi-square and adaboost on one of the classification algorithms, namely C5.0. In this study, the majority class in the dataset used is dominated by the negative class, so the performance appraisal should focus more on the positive class. Therefore, a more suitable assessment is recall/sensitivity/TPR because the resulting value only depends on the positive class. The results showed that both methods were able to increase the recall/sensitivity/TPR value, meaning that the application of chi-square and adaboost was able to improve the classification performance of the minority class