Predictive modeling of students performance through the enhanced decision tree

Abstract

Prognostic of student performance is one of the major issues in many institutions. The student’s performance is based on many factors such as internal examinations, grade obtained in university examination, Academic, Extra Curricular and Co-Curricular activities and also concern with their activities in learning environment. Student performance prediction is used to model the students into any one of the four categories as excellent, good, average, and poor performance student. The category selection was determined using supervised classifiers. Academic institution can easily able to identify any academic problems and the corresponding solutions for their students through this predictive modeling approach. We have collected real world dataset related to student’s academic performance from leading academic institution in India which consists of details about the students such as CGPA, Lab performance, History of arrears and so on. We have applied various supervised classifiers such as DT, SVM, KNN, NB, NN and Improved DT on student’s academic performance dataset. Besides, the conventional decision tree is further improved by the use of normalized factor and Association Function. By comparing the accuracy of these two methods, the best result is exposed for Improved Decision Tree than all other supervised classifiers in the literature.