EDUCATIONAL DATA MINING METHODS FOR TIMSS 2015 MATHEMATICS SUCCESS: TURKEY CASE

Abstract

Educational data mining (EDM) is an important research area which has an ability of analyzing and modeling educational data. Obtained outputs from EDM help researchers and education planners understand and revise the systematic problems of current educational strategies. This study deals with an important international study, namely Trends International Mathematics and Science Study (TIMSS). EDM methods are applied to last released TIMSS 2015 8th grade Turkish students' data. The study has mainly twofold: to find best performer algorithm(s) for classifying students' mathematic success and to extract important features on success. The most appropriate algorithm is found as logistic regression and also support vector machines - polynomial kernel and support vector machines - Pearson VII function-based universal kernel give similar performances with logistic regression. Different feature selection methods are used in order to extract the most effective features in classification among all features in the original dataset. Home Educational Resources, Student Confident in Mathematics and Mathematics Achievement Too Low for Estimation are found the most important features in all feature selection methods.