ETL Preprocessing with Multiple Data Sources for Academic Data Analysis

Abstract

More and more, the needs of academic data analysis are requiring in educational settings for the purpose of improving student learning and institutional effectiveness. In the education world, testing and data are driving decisions for what knowledge and skills students should be learning and how students’ learning relates to their learning outcomes. On that occasion, a better option for building a machine learning model is to get effective data preprocessing concepts. For these reasons, this paper describes the very first step of the main research work which considers the correlations between students’ academic performance, behavior and personality traits to reveal the presence of an intriguing way. Intuitively, this paper proposes the uses of Extraction-Transformation-Loading (ETL) in the preprocessing stage to collect and analyze of students’ data from multiple data sources. In this system, data is collected from multiple data sources based on the structures which are used as a testbed. Students’ demographic data and assessment results from Student Information System (SIS), logs of their interaction with Moodle are used for data collection. Then aggregating with Web logs also captures student behavior that is represented by daily summaries of student clicks based on courses and by their actions.