PENGGUNAAN ALGORITMA CLASIFICATION AND REGRESSION TREE (CART) UNTUK MENDUKUNG KEPUTUSAN PADA PENILAIAN SERTIFIKASI GURU

Abstract

ABSTRACTPortfolio assessment in teacher certification is essentially a form of competency test to obtain acertificate of educators. Therefore the assessment ofteacher portfolios as limited as an assessment of thephysical evidence collection that reflects theachievements of teachers track record in carrying outhis duties as a basis for determining the level ofprofessionalism of teachers is concerned. Portfolioassessment of each participant teacher certificationper-formed by two assessors, and the componentsare assessed stored in a database Aplikasi Sertifikasi
Guru (ASG), which until now has not beenused for more detailed assessment. This research will use algorithmsClassification and Regression Tree (CART) whichis one of the methods or algorithms of any of thetechniques of data exploration, namely the decisiontree technique. This method was developed by LeoBreiman, Jerome H. Friedman, Richard A. Olshenand Charles J. Stone circa 1980's. CART is anonparametric statistical methodology developed fortopic classification ana-lysis for both continuous andcate-gorical response variables. CART produces aclassification tree if the response was categoricalvariables, and produce regression tree if the responsevariable is continuous. CART main goal is to get agroup of data is accurate as identifier of theclassification. And typical CART algorithm is adecision node always forked or branched binary. The data used for this research is theimplementation of a data certi-fication of teachers in2009 and 2010, in which case the final determinationof a portfolio that is different between the twoassesor, the first assesor declared pass while thesecond assessor states did not pass or take part ineducation and training of the teaching profession(PLPG), there were 236 participants. By setting 80%for training data and 20% for random testing of data, 66the value of 80.31% accuracy, and the resulting treeis: