Open AccessUsing Predictive Modeling System and Ensemble Method to Ameliorate Classification Accuracy in EDM
Author(s) -
Mudasir Ashraf,
Majid Zaman,
Muheet Ahmed
Publication year - 2018
Publication title -
asian journal of computer science and technology
Language(s) - English
Resource type - Journals
eISSN - 2583-7907
pISSN - 2249-0701
DOI - 10.51983/ajcst-2018.7.2.1874
Subject(s) - undersampling , c4.5 algorithm , boosting (machine learning) , oversampling , computer science , ensemble learning , machine learning , decision tree , random forest , naive bayes classifier , artificial intelligence , random subspace method , support vector machine , data mining , computer network , bandwidth (computing)
Educational data mining has illustrated an increasing demand for extracting and maneuvering data from academic backdrop, to generate prolific information which is indispensible for decision making. Therefore in this paper, an attempt has been made to deploy various data mining techniques including base and meta learning classifiers across our pedagogical dataset to foretell the performance of students. Among several contemporary ensemble approaches, researchers have practiced widespread learning classifiers viz. boosting to predict the performance of students. As exploitation of ensemble methods is considered to be significant phenomenon in classification and prediction mechanisms, therefore analogous method (boosting) has been applied across our pedagogical dataset. The entire results have been evaluated with 10-fold cross validation, once pedagogical dataset has been subjected to base classifiers including j48, random tree, naive bayes and knn. In addition, techniques such as oversampling (SMOTE) and undersampling (Spread subsampling) have been employed to further draw a comparison among ensemble classifiers and base classifiers. These methods were exploited with the key objective to observe any improvement in prediction accuracy of students.