Multivariate Adaptive Regression Splines

Abstract

Education is a top priority for today's society. The quality of education can be seen from the learning achievement. There are so many factors that influence learning achievement in this regard graduation, therefore, necessary to identify the most influential factors that will be used to improve the quality of education. This study was conducted to obtain a model that is capable of classifying the data Faculty of Science and Mathematics Diponegoro University Semarang graduation using Multivariate Adaptive Regression Spline (MARS) method. MARS is a nonparametric regression method that can be used for data of high dimension. To get the best MARS models, made possible combinations Basis Function (BF), Maximum Interaction (MI), and Minimum Observation (MO) by trial and error. The best model is the model that is used in combination with BF = 28, MI = 2, MO = 1 because it has the smallest GCV value that is equal to 0,17781. There are three variables that contribute to the MARS model of the variable GPA, majors and gender. As for the variable organization, part time, entry point, and scholarships do not contribute to the model. Obtained misclassification of 20,50%. Press's Q test value indicates that statistically MARS method has been consistent in classifying the data FSM Diponegoro University Semarang graduation.