Pedagogical Data Federation toward Education 4.0

Abstract

Pedagogical data analysis has been recognized as one of the most important issues in pursuing Education 4.0. The recent rapid development of IT technologies benefits pedagogical data analysis via the provisioning of many advanced technologies such as big data analysis and machine learning. Meanwhile, the privacy of the students become another concern and this makes the educational institutions reluctant to share their students' data, forming isolated data islands and hindering the realization of big pedagogical data analysis. To tackle such challenge, in this paper, we propose a federated learning based education data analysis framework FEEDAN, via which pedagogical data federations can be formed by a number of institutions. None of them needs to direct exchange the their pedagogical data with each other and they always keep the data in their own place to guarantee their students' privacy. We apply our framework to analyze real pedagogical data. The experiment results show that it not only guarantees the students' privacy but also indeed breaks the borders of data island by achieving a higher analysis quality. Our framework can much approach the performance of centralized analysis which needs to collect the data in a common place with the risk of privacy exposure.