Synthetic minority over-sampling technique nominal continous logistic regression for imbalanced data

Abstract

Evaluation is a method that determines the success of a program or policy that has been carried out by an institution,one of which is an educational institution. Service evaluation is carried out by means of a satisfaction survey of students. The resultsof the evaluation are one of the efforts to improve the quality of services in educational institutions. In 2021 in one of the educationalinstitutions in Surabaya, 5% of students were dissatisfied and 95% of students were satisfied. These data indicate an imbalance in the classification of student satisfaction, where the number of students in the satisfied category dominates more than the number ofstudents who are dissatisfied. The data imbalance has a negative impact on the classification where the minority class is often misclassified. Therefore, this study classifies student satisfaction using the Logistics Regression and the Synthetic Minority OverSampling Technique Nominal Continuous (SMOTE-NC) Logistic Regression. The results of the study using the Logistics Regression have an accuracy of 79.41%, a sensitivity of 84%, a precision of 93% and an error rate of 21%. The SMOTE-NC Logistics Regression has 85.29% accuracy, 91% sensitivity, 94% precision and 15% error rate. The results of this study concludedthat the SMOTE-NC Logistics Regression method can handle the imbalance in the amount of data and has better accuracy, sensitivity, precision and error rate values than Logistics Regression.