Perbandingan model klasifikasi linear discriminant analysis dan K-nearest neighbor untuk data penjurusan siswa Madrasah Aliyah Negeri Samarinda

Abstract

Madrasah Aliyah Negeri (MAN) under Ministry of Religion used curriculum 2013 as well as others senior high schools. However, MAN subjects are more than others senior high schools. It’s became particular concern to parents of students so that their children can compete with graduates of public schools when they continue to higher education. Curriculum 2013 provides that majoring in Ilmu Pengetahuan Alam (IPA), Ilmu Pengetahuan Sosial (IPS), Agama, and Bahasa have been implemented since the 10th grade. Picking the wrong major can be avoided by knowing the characteristics of students. The process of finding data patterns for student majors can be done using data mining. The purpose of this research was to classify the data of placement MAN’s student majors using Linear Discriminant Analysis (LDA) model as linear model and k-Nearest Neighbor (k-NN) model as non-linear model. The data were resampled using Bootstrap with n=1000 and n=5000. The data were also divided into training data and testing data with the probability of each data being drawn was 60:40, 70:30, 80:20, and 90:10. Based on the results, the accuracy of LDA models for distribution of training data and testing the data at 60:40, 70:30, 80:20, and 90:10 and bootstrap n=5000 respectively were 0.838, 0.834, 0.833, 0.832. Meanwhile the accuracy of k-NN models with k=3 for distribution of training data and testing the data at 60:40, 70:30, 80:20, and 90:10 and bootstrap n=5000 respectively were 0.837, 0.846, 0.854, 0.861. Therefore, the k-NN model with k=3 became the best model.