Predicting and analyzing secondary education placement-test scores

摘要: Highlights? Understanding the success factors for placement tests is a challenging problem. ? Analysis of these may be used to improve test structure and content. Data mining can effectively model analyze scores. Results showed that C5 decision tree algorithm best predictor with 95% accuracy. Sensitivity analysis revealed previous experience most important predictor. lead (or failure) students at an interesting Since centralized future academic achievements are considered related concepts, behind help understand potentially achievement. In this study using large feature rich dataset from Secondary Education Transition System in Turkey we developed models predict secondary education results, sensitivity on those prediction identified predictors. The results accuracy hold-out sample, followed by support vector machines (with 91%) artificial neural networks 89%). Logistic regression came out least accurate four overall 82%. experience, whether student has scholarship, student's number siblings, years' grade point average among predictors