Exploring and Identifying Student Engagement and Performance Profiles in A Learning Environment

Abstract

Many studies illustrate the potential of utilizing data and advanced analytics techniques in specific and particular educational and learning settings. Much less focus is devoted to applying broader and more general approaches to exploratory data analysis in the field of Learning Analytics (LA). This thesis presents an additional contribution in this space. It demonstrates a general approach to exploring and predicting students’ profiles in an online learning setting. Its intention is to contribute to the growing literature of exploratory research in the field of LA by applying robust approaches to analysing data about students’ behaviours. The thesis describes a case study where data were collected via a naturalistic experiment relating to an actual, real-life group of students from a first semester engineering course. Subjects in this naturalistic experiment were engaging with a learning content through a learning management system (LMS) in a blended-learning environment. The approach in this thesis leverages modern advanced analytical techniques to contribute to the growing body of literature in the field of exploratory analysis in LA. A combination of unsupervised and supervised statistical learning methods are applied to identify, cluster and, subsequently, classify groups of students based on their profiles. The results suggest the existence of distinct groups of students with fairly distinguishable characteristics. Findings also illustrate the application of predictive analytics models to classify students based on their previously identified characteristics. The thesis concludes with discussions on potential implications and limitations which intend, ultimately, to help researchers, educators and institutions alike understand, build and deliver a more adaptive learning experience and environment from an exploratory data analysis perspective.