Bootstrap Aggregating Multivariate Adaptive Regression Splines

(Studi Kasus : Identifikasi Komponen Penciri Akreditasi Sekolah/Madrasah Pada Tingkat SD/MI di Provinsi Kalimantan Timur Tahun 2015)

Abstract

MARS is one of the classification methods that focus on the high dimension and discontinuity of the data. The level of accuracy in MARS can be improved by using Bagging method (Bootstrap Agregating). This method is used to improve stability, accuracy and strength’s of prediction. This study discusses the MARS bagging applications in analyzing the issue of accreditation, which the accreditation level of a schools can be predicted based on the identifier components. Therefore, in this study will be identified these components to create a classification model. The data used is the accreditation data of the primary school in East Kalimantan Province 2015 issued by the Accreditation Board of the Provincial Schools (BAP-S/M) of East Kalimantan Province. This study obtained six components that affect the determination of the accreditation of schools at primary school level. The components are the variables that contribute to the classification. The variables are a standard component of content (X₁), a standard component of the process (X₂), a standard component of graduates (X₃), standard components of teachers and staffs (X₄), a standard component of infrastructure (X₅) and standard component of financial (X₇). Based on the result of the classification accuracy of MARS method (using Apparent Error Rate (APER), it is amounted to 78.87%, while the classification accuracy (using APER) with method of bagging of the best MARS models amounted to 89.44%. This means that the method of bagging MARS gives better classification accuracy of the classification than MARS.