Using Data Mining Approach to Predict Collegian’ Improvement Study

Understanding the factors that lead to success of university students’ education is aninteresting subject. For our research we have chosen two universities in Qom city in the centerof Iran. Scores obtained from average taken by university students are combined with otherfactors using a preset formula to determine the next term average .The aim of this study istwofold: (i) to investigate predictive power of C۵ & SVM methods and (ii) to determine theranked-importance of predictive variables by applying sensitivity analysis on predictionmodels . The results showed that C۵ decision tree algorithm is the prediction of ۷۷.۳۹%accuracy on ۱۰-fold holdout dataset. Decision tree models are followed by support vectormachines with an overall prediction accuracy of ۶۲.۱۱%.

Keywords:

Data Mining , Decision tress , Support Vector Machine , Predicting student performance