Performance evaluation of classifiers for edu-data: an integrated approach

This paper mainly focuses on the performance evaluation of the three types of classifiers viz., Rule-based, Decision tree based and Baysian networks on Edu-data which is a large repository consisting of data related to technical education system; which is considered as a benchmark system for the study of Edu-mining. The three important stakeholders of the system are Student, Faculty and Management. The study comprises of 3 modules in student stakeholder, 3 modules in faculty stakeholder and finally an integrated module. Totally 3500 instances are taken for each module and the results of the present investigation predict (i) the optimal classifiers for each module.(ii) the accuracy and time complexity for all the classifiers and (iii)facilities to take the effective managerial decisions. No doubt the results of the present integrated approach provides a unique platform for making effective predictions at all the levels.