Al-Jumeily, R (2024) Implementation of an Automated Recommender System for Classification and Prediction of Students’ Performance in a Multi-Educational Systems Using Machine Learning Analytics. Doctoral thesis, Liverpool John Moores University.

Text 2024Al-JumeilyPhD.pdf - Published Version Restricted to Repository staff only until 9 September 2026. Available under License Creative Commons Attribution Non-commercial. Download (5MB)

Abstract

Aim: To develop and validate an automated recommender system (ARS) for predicting students’ performance in a multicultural and diverse environment using artificial intelligent (AI) and machine learning analytics. Method: The system was developed and proposed in a school in Abu Dhabi that had students from different backgrounds. The school was a British School that had 950 students annually between the age groups of 4-16 years old. A multimodal approach was used that comprised action research, quantitative statistical methods and AI and machine learning was used over three phases. Phase I involved implementation of an educational robot in maths and automated recommender system for education. The usability of the robot was assessed by a questionnaire and the recommender system using neural networks and deep-learning-based models. Phase II comprised unsupervised machine learning analytics for classifying students’ performance and identifying key features contributing to performance. These analytics included correlation methods, principal component analysis and self-organising maps. Phase III encompassed supervised machine learning models for early prediction of students’ performance. The latter analytics included principal component regression and partial least square regression. It is worth noting that phases II and III used 11 features obtained from the ARS used in phase I. All statistical and machine learning analytics were conducted using Matlab 2019a. For evaluation of the system, teachers gave feedback on the usability of the ARS and its content. In addition, validation and verification of the ARS was conducted by two groups: the first included teachers, and the second comprised expert panel of teachers and data scientists. Findings: Abu Dhabi has 14 different educational systems that include, but are not limited to, a British, American and Asian systems. Students often move between these educational systems and teachers encounter challenges in levelling up students coming from a different system. Therefore, this research proposed, evaluated, and validated an ARS for levelling up students. The results showed the effectiveness of the approach in the multicultural/multi-educational environment proposed. Hence, in phase I students and teachers gave positive feedback on the usability of the robot and the ARS. Moreover, neural networks and deep-learning based models applied in phase I showed high accuracy in measuring students’ performance. Moreover, in phase II unsupervised analytics applied to the ARS proved efficient in classifying weak and well-performing students, as well as identifying key features contributing to optimal performance that were related to attendance, practice exercises and homework. Phase III also proposed a system of early prediction of performance based on features obtained from the automated recommender system. In this respect, the two supervised machine learning analytics showed high accuracy and precision of prediction. Conclusions: The innovative approach proposed in this research showed effectiveness not only in classifying and predicting students’ performance in maths but also in identifying key features that contribute to performance. It did not show any limitations that are usually encountered in traditional teaching and learning approaches, especially in a multicultural environment as in the present study. The findings highlighted the urgent need to improve educational procedures relating to delivery and assessment in order to facilitate educational culture, provide equitable access to learning for all and reduce educational inequalities at a global level.

Item Type:	Thesis (Doctoral)
Uncontrolled Keywords:	automated recommender system; education; students performance; machine learning analytics; classification; prediction
Subjects:	Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Divisions:	Computer Science & Mathematics
SWORD Depositor:	A Symplectic
Date Deposited:	10 Sep 2024 14:06
Last Modified:	10 Sep 2024 14:06
DOI or ID number:	10.24377/LJMU.t.00022727
Supervisors:	Kolivand, H, Hussein, A and Khan, W
URI:	https://researchonline.ljmu.ac.uk/id/eprint/22727

View Item