Perbandingan Algoritma Klasifikasi Data Mining untuk Penelusuran Minat Calon Mahasiswa Baru

Abstract

During the AMIK HASS pandemic, it was difficult to determine new student candidates. So that to attract public interest, the Marketing Department has implemented several strategies to attract prospective students to become new students. The data mining technique used in predicting is a classification that includes Naïve Bayes, J48 Decision Tree, and K-Nearest Neighbor. This study aims to perform a comparative analysis of data mining classification algorithms using WEKA tools. The method used in this study, using CRISP-DM. The dataset used by the three classifications is 5,934 records with split mode, the percentage of testing is 70% as much as 4154 as training data and 30% as much as 1780 data as test data. Based on the test results on the three classification models, the highest accuracy value is obtained in the J48 Decision Tree classification, which has a value of 90.3%. While the K-Nearest Neighbor classification has a lower accuracy of 87.52% and the Naïve Bayes classification has an accuracy of 87.24%. The comparison of the AUROC J48 Decision Tree test results has the highest value of 0.9654 while the Naïve Bayes results are 0.9461 and the K-Nearest Neighbor results are 0.9343. The three classifications with ABK scores above 0.90 are included in the excellent classification category