Towards enhancement of privacy-preserving data mining model for predicting students learning outcomes performance

Abstract

There is massive data collection such as the student's information which is included their profile background and learning outcome performance that had been collected by the academic institution. With those of the data collection to identify the potential students who can contribute towards the development of the academic institution. An academic institution that provides open and distant learning programs can advantage from big data on their students' information, as well as data mining techniques and big data analytics tools, provided well as the students' right to privacy is protected. However, the student records shown the potential for enhanced information sharing throughout academic institution departments which raised up the privacy concern. Therefore, this study proposes the enhancement of Privacy Preserving Data Mining (PPDM) model for predicting students' learning outcomes performance strikes a better accuracy of data mining while preserving better data privacy. This PPDM combines two privacy preserving approaches which are k-anonymization and homomorphic encryption. On the other hand, for the data mining, classification algorithms such as Random Forest, Support Vector Machine, and Naïve Bayes are compared to find the best accuracy. The outcome of the proposed PPDM model is expected to give better prediction accuracy for students learning outcomes performance as well as high privacy preservation.