Development of important patterns in student database using statistical classifier

Abstract

The research aims to use some data mining techniques to predict the phenomenon of leakage of undergraduate students using a number of risk factors (important) ( the gender , the attendance , the former grade of students , the educational level of parents , friends , first child , working ) by using the closet nearest neighbor algorithm (KNN). 350 survey forms were distributed among the students of all the stages of the college of administration and economy department of administrational information systems. The forms contained 19 questions. The variables of the forms and their data took the classification of (C4.5) and CART to compare the results and to choose the best to predict the rate of success and failure of the students who have failed before . It has also been designed a database for the students of college of Administration & economics , with a computer system related on concerning with all registration affairs.