Implementation Of Data Mining In Determining Scholarship Eligibility At SMA Kristen Tunas Kasih

Abstract

Determining the eligibility of scholarship recipients often faces problems of subjectivity and inaccuracy of targeting due to the limitations of manual assessment. This research implements the Modified K-Nearest Neighbor (MKNN) method to determine scholarship eligibility at SMA Kristen Tunas Kasih, under the auspices of the Gereja Kebagunan Kalam Allah Indonesia (GKKAI). The goal is to build an accurate and objective classification system for selecting financially disadvantaged yet promising students. The dataset consists of 112 active students from the 2023/2024 academic year, divided into 90 training data and 22 testing data (80:20 ratio). Classification criteria include parental employment, parental income, and parental status. The MKNN method is implemented through several stages: Euclidean distance calculation, data normalization, validation, and weight voting. The system is developed as a web application using PHP and MySQL, featuring student data management, training-testing data division, weight calculation, and result visualization. Testing results show an accuracy of 95.45%, precision of 100%, recall of 75%, and F-measure of 85.71%. Out of 22 testing data, 21 were classified correctly, with only 1 mismatch. The optimal K value found through Grid Search is K = 3, yielding the best classification. This study demonstrates that the MKNN method delivers strong performance in determining scholarship recipient eligibility.