Sistem Klasifikasi Kelayakan Penerima Beasiswa menggunakan Algoritma Categorical Naive Bayes

Abstract

Scholarship selection is often done manually, which takes time and has the potential for errors in assessment. This study aims to build a prediction model for student scholarship eligibility using the Categorical Naive Bayes algorithm. The data used in this study consisted of 1,042 student data with eight attributes: distance from residence to campus, gender, organizational participation, student activity unit (UKM) participation, GPA, parents' occupation, parents' income, and number of dependents. The research method included data preprocessing, feature encoding, data splitting with an 80:20 ratio, model training using three Naive Bayes variants (GaussianNB, CategoricalNB, and ComplementNB), and model evaluation using accuracy, precision, recall, F1-score, and AUC-ROC metrics. The results showed that the Categorical Naive Bayes model achieved the best performance with an accuracy of 74.16%, precision of 50.85%, recall of 54.55%, F1-score of 52.63%, and AUC-ROC of 80.25%. The most influential features in determining scholarship eligibility were the number of dependents, GPA, and organizational participation. This study concludes that the Categorical Naive Bayes algorithm can be used to predict scholarship eligibility reasonably well, although it still needs improvement to handle imbalanced data.