GDL

Seleksi Nasional Masuk Perguruan Tinggi Negeri (SNMPTN) is one of the selection pathways for student admissions to enter state universities (PTN) in Indonesia. This study aims to predict the chance of being accepted in the desired PTN and the lack of early monitoring of students for SNMPTN. The data source from the grades reports card of SMAN 1 Pakong, SMAN 8 Kediri, and SMAN 1 Pamekasan by using the average input of compulsory subjects, majors (Science / Social Sciences) and semester 1 to semester 5 which later the output to be accepted or not accepted An imbalanced dataset potentially affect the performance of the classification method used. Hence, we need to eliminate the imbalance class using SMOTE. Using 10-fold cross validation, this study compared K-Nearest Neighbor (KNN) without SMOTE and K-NN with SMOTE. The goal is to find the best prediction model between the two methods. The prediction model is applied to software for teachers to monitor student grades and ensuring students to pass the SNMPTN. The results show that KNN without SMOTE has higher accuracy than KNN with SMOTE. However, KNN with SMOTE outperform than KNN without SMOTE in precision and recall, KNN with SMOTE with K = 3 reached 80.08% Accuracy, 74.42% Precision and 91.68% Recall.

Beri Komentar ?#(0) | Bookmark