Applicability of Traditional Classification Techniques on Educational Data
Balwinder Kaur1, Anu Gupta2, R. K. Singla3
1Balwinder Kaur*, Defense Security Cooperation Agency, Panjab University, Chandigarh, India.
2Aun Gupta, Defense Security Cooperation Agency, Panjab University, Chandigarh, India.
3Ravinder K. Singla, Defense Security Cooperation Agency, Panjab University, Chandigarh, India.
Manuscript received on February 10, 2020. | Revised Manuscript received on February 20, 2020. | Manuscript published on March 30, 2020. | PP: 1672-1677 | Volume-8 Issue-6, March 2020. | Retrieval Number: E6149018520/2020©BEIESP | DOI: 10.35940/ijrte.E6149.038620
Open Access | Ethics and Policies | Cite | Mendeley
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: Student performance prediction and analysis is an essential part of higher educational institutions, which helps in overall betterment of the educational system. Various traditional Data Mining (DM) techniques like Regression, Classification, etc. are prominently utilized for analyzing the data coming from educational settings. The usage of DM in the area of academics is called Educational Data Mining (EDM). The current pilot study aims to determine the applicability of these standalone classification techniques namely; Decision Tree, Bayes Net, Nearest Neighbor, Rule-Based, and Random Forest (RF). The present pilot study uses the WEKA tool to implement traditional classification techniques on a standard dataset containing student academic information and background. The paper also implements feature selection to identify the high influential features from the dataset. It helps in reducing the dimensionality of the dataset as well as enhancing the accuracy of the classifier. The results of classifiers are compared on basis of standard statistical measures like Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Kappa, etc. The results show the applicability of classification algorithms for student performance prediction which will help under-achievers and struggling students to improve. It is found the output that, J48 algorithm of the Decision tree gave the best results. Further, it is deduced from the comparative analysis that individual classifiers give different accuracy on the same dataset due to class imbalance in a multiclass dataset.
Keywords: Performance Prediction, Data Mining Techniques, Classification, Decision Tree, BayesNet, Random Forest, WEKA.
Scope of the Article: Classification.