| dc.description.abstract | Classification tree is a simple method to determine the factors that influence a response variable whose type is categorical. However, this method has some problems. First, selection bias on explanatory variables (called selection bias ) and bias due to missing value . This problem can be solved by CRUISE (Classification Rule with Unbiased Interaction Selection and Estimation) 2D. Another problem is the imbalance of data, so that observations in the minority class (a class that have less observation’s number) can’t be classified correctly. SMOTE (Synthetic Minority-Over Sampling Technique) with majority under-sampling is a method that can be used to handle those problems. The aim of this study is determining academic success factors of BUD (Beasiswa Unit Daerah) students, which has majority class (‘non-drop out’ category) and minority class (‘drop out’ category). The classification capability of CRUISE 2D tree without SMOTE is very bad, so it’s better to choose the tree from SMOTE with majority under-sampling which use 300 percent of increasing and 133.34 percent of decreasing. This tree has 12 terminals and 5 of them have high risk to be dropped out. Variables that influence the academic result of BUD students are the origin island, the average value of the national exam, the father and mother's education, sponsorship type, parent’s income, and mother’s education. About 73.7% observations with DO class in testing data can be classified correctly by this tree. | en |
