CORRELATION BASED FEATURE SELECTION (CFS) TECHNIQUE TO PREDICT STUDENT PERFROMANCE

Education data mining is an emerging stream which helps in mining academic data for solving various types of problems. One of the problems is the selection of a proper academic track. The admission of a student in engineering college depends on many factors. In this paper we have tried to implement a classification technique to assist students in predicting their success in admission in an engineering stream.We have analyzed the data set containing information about student’s academic as well as sociodemographic variables, with attributes such as family pressure, interest, gender, XII marks and CET rank in entrance examinations and historical data of previous batch of students. Feature selection is a process for removing irrelevant and redundant features which will help improve the predictive accuracy of classifiers. In this paper first we have used feature selection attribute algorithms Chi-square.InfoGain, and GainRatio to predict the relevant features. Then we have applied fast correlation base filter on given features. Later classification is done using NBTree, MultilayerPerceptron, NaiveBayes and Instance based –K- nearest neighbor. Results showed reduction in computational cost and time and increase in predictive accuracy for the student model