A Comprehensive Analysis on Undergraduate Student Academic Performance using Feature Selection Techniques on Classification Algorithms

Abstract

Educational Data Mining (EDM) is a growing research field that is applied to analyze and predict student's academic performance and makes intervention approaches to elevate that performance. It is a field of study, which is related to various attributes for analysis student's details such as name, attendance, class test, lab test, spot test, assignment and result in the educational institution. In this study, we mainly focus on calculating the academic performance of undergraduate students with a predictive data mining model by using feature selection techniques with classification algorithms. Feature selection techniques are introduced on the data preprocessing process to find the most inherent and important attributes so that we analyze and evaluate the student's better performance by using classifiers with those selected attributes. For this purpose, we collected 800 student's records of the final year, studying at the undergraduate level of the department of Computer Science and Engineering from North Western University, Khulna. Here, we used and evaluated the performance of four feature selection methods: genetic algorithms, gain ratio, relief, and information gain and five classification algorithms: K-Nearest Neighbor, Naïve Bayes, Bagging, Random forest, and J48 Decision Tree. The experimental results depict that Genetic algorithms method provides the best accuracy 91.37% with KNN classifier.