Data Sharing Technique Modeling for Naive Bayes Classifier for Eligibility Classification of Recipient Students in the Smart Indonesia Program

Abstract

The objective of Smart Indonesia Program (Program Indonesia Pintar: PIP) is to help school-aged people from poor/vulnerable/priority families to continue to receive education services to graduate from secondary education, both through formal and non-formal education channels. In its implementation, there are still many fraudulent in the proces of nominating proposal PIP funds and there are still many prospective students who should not receive PIP because they do not meet the technical guidelines provided by the Ministry of Education and Culture to determine the eligibility of prospective recipients of PIP funds can be done by schools and stakeholders, one of them by using classification techniques. One algorithm that is widely used in classification is the Naive Bayes Classifier (NBC) algorithm. In this study three data sharing techniques were used, namely Hold Out 70% training data and 30% testing data, K-Means Clustering, and also 10 Fold Cross Validation. Determination of the best data sharing technique will be determined by looking at the value of Accuracy, Precision, and Recall and also the value of Area Under Curve (AUC) which is illustrated by the Receiver Operating Characteristic (ROC) curve so that the NBC algorithm is generated with 10 Fold Cross Validation has a very good classification level with the values of accuracy, precision, and recall respectively at 97.40%; 100%; and 76.14%.