DIGITAL LIBRARY
REDUCING LEARNER ATTRITION THROUGH AI BASED PREDICTIVE MODELING
1 Smart Apprentices (UNITED KINGDOM)
2 Aston University (UNITED KINGDOM)
About this paper:
Appears in: ICERI2021 Proceedings
Publication year: 2021
Pages: 440-444
ISBN: 978-84-09-34549-6
ISSN: 2340-1095
doi: 10.21125/iceri.2021.0164
Conference name: 14th annual International Conference of Education, Research and Innovation
Dates: 8-9 November, 2021
Location: Online Conference
Abstract:
Digitalisation and automation are ever growing topics, with expanding application domains and accelerating technological innovation. Artificial Intelligence (AI) and Machine Learning (ML) are driving the world into a future of augmented decision making with faster, better and more consistent predictions and performance, helping enhance human development in all domains varying from fishing in the oceans to space exploration. One field where the application of AI is still incipient is the education industry, where the potential to revolutionise existing systems through the automation of mundane tasks that occupy the time of education professionals remains to be realised. AI can also be used to introduce personalised, data driven decision making for better student learning journeys. A starting point for the development of such tools is a system capable of understanding learner progress and predicting which ones may struggle to complete their studies months in advanced, based on learner-specific data and similarities across historical learner outcomes. We propose the Smart Coach concept, an ML framework offering a complete pipeline to mine relevant data for student performance prediction, including relevant feature extraction and importance ranking. Using the processed data, glass box AI modelling tools are applied for interpretable learner outcome prediction. The proposed predictive model can help automatically identify struggling students, allowing education providers to focus on designing intervention procedures to help successfully complete their learning journey. These decisions can also be collected and augmented with the framework to produce a decision system for automated interventions, with the goal of providing a more robust learning environment where learners are offered personalised guidance in times of struggle. The design of a full interpretable framework is important for the tractable analysis and expert-driven pattern identification, which is needed for further perspective analysis and personalised intervention procedure design. We demonstrate the results on data collected by Smart Apprentices’ Smart Assessor, an e-portfolio system for apprenticeship education in the UK, where high predictive accuracy is successfully achieved.
Keywords:
Artificial intelligence, machine learning, future of learning.