EXPLORING THE USE OF MACHINE LEARNING TO IMPROVE STUDENT ENGAGEMENT AND RETENTION I. Ogbuchi, E. Kiely, C. Quigley, D. McGinty Atlantic Technological University (IRELAND) Abstract This paper reports on the analysis of Moodle (an open-source online learning platform) data from first- year computing students (n=200) participating in a Maths module, the 2020/21 data revealed ~ 213,000 interactions with the Virtual Learning Environment (VLE). These interactions were analysed and showed a strong correlation with student performance. Machine learning models were explored as a means of clustering students by their level of engagement. Retention issues relating to the risk of disengagement is analysed in relation to interactions and academic performance. This paper begins with a brief introduction of learning technologies currently used to evaluate student engagement an Irish University Context. Data is extracted from Moodle into a database where it is cleaned and structured. The analysis identified engagement trends in the student data which was explored by variables such as day, time, and activity type. Risk categories and thresholds are defined by last access to VLE, academic performance and engagement with specific module. The model correctly identified students who failed the module as early as week 5. In this study, students are partners in the design process and feedback of their experiences was collected. The students are also introduced to explainable machine learning technologies and the steps in the analysis of the study. This provides a constructive alignment between our research aims, achievement of programme learning outcomes (as computing students) and enhancement of the student experience. Keywords: Moodle. 1 INTRODUCTION With the global adoption of Virtual Learning Environments (VLEs) as an integral component of learning in universities, big data on how students are engaging is generated (Drachsler & Greller, 2012) [1]. Daud et al. (2017) noted in his work that ability to predict whether a student will be successful is of value [2] as it can aid institutions to make quick decisions about their students before they decide to drop out by offering additional support or more adapted learning interventions. Machine learning analytics can be applied to complement this by highlighting patterns and predictions of student behavior. The primary goal of analytics according to Chatti et al. (2012) is exploring the value of data gathered in providing learning community (including students) with actionable information that could be used to enhance the learning environment [3]. Mirriahi et al., (2014) proposes that the application of data analytics in educational settings is a rapidly growing research discipline [4] which can lead to insights and make predictions of students’ performance and inform retention strategies. Machine learning (ML) is a branch of artificial intelligence (AI) and computer science which focuses on the use of data and algorithms to imitate the way that humans learn, gradually improving its accuracy [5]. This holds a lot of value for big data generated by educational institutions which is often not analyzed for insights. Machine learning learns from data fed into it using programming and mathematical tools to learn patterns in data which are often difficult for the academic community to see. Machine Learning types include supervised, unsupervised, reinforcement learning, semi supervised and learning to learn [6]. Supervised learning techniques focus on teaching the computer how to learn from data by showing it what to look for while in unsupervised learning the model automatically looks for the patterns in the data without necessarily being shown what to look for. Reinforcement learning is when the model is trained using a reward system whereby it is acts on an environment and is penalized or rewarded according to a set of actions it takes. Proceedings of ICERI2022 Conference 7th-9th November 2022 ISBN: 978-84-09-45476-1 3385