Sistem Rekomendasi Bidang Program Studi Menggunakan Algoritma K-Nearest Neighbor Berbasis Web

Abstract

Selecting a suitable study field is a recurring challenge for final-year senior high school students because the choice affects their academic path and career orientation. At SMA Negeri 5 Tangerang Selatan, counseling activities require a data-supported instrument that can provide an initial view of students’ possible study-field tendencies. This research designed a web application that applies the K-Nearest Neighbor approach to compare student academic scores with alumni records that already contain known study-field labels. The application was built through the Waterfall development stages using Python, Flask, SQLite, HTML, CSS, and JavaScript. The research data comprised 450 alumni records with four academic attributes: Mathematics, Indonesian Language, English, and Science/Social Studies. Classification was performed through data preparation, Min-Max scaling, distance measurement with Euclidean Distance, selection of nearest neighbors, and majority voting. The algorithm was tested using 10-fold cross validation and a confusion matrix for K values of 1, 3, 5, 7, 9, 11, 13, and 15. The highest result appeared at K=11, with 25.11% accuracy, 25.21% precision, 24.65% recall, and 24.68% F1-score. These findings indicate that the application is more appropriate as supporting information for counseling discussions than as a sole basis for determining study choices.