All Theses and Dissertations


Doctor of Philosophy in Computer Science


Department of Computer Science

Date of Award

Spring 2017


Dr. Shakeel Ahmed Khoja

Committee Member 1

Dr. M. Ashraf Iqbal, University of Central Punjab, Lahore, Pakistan

Committee Member 2

Dr. Nelofer Halai, Aga Khan University, Karachi, Pakistan

Project Type


Access Type

Restricted Access


xiv, 195


The inherent complexity of online learning environments and the abundance of resources in these environments cause present-day learners information-overload, knowledge-deficit, and personalisation-related challenges. Learning Environment or PLE tackles these issues with its integral user-centred desian.. The concept of on Personal recommendation networking and persistent online presence features. This thesis presents Guided PLE Model, a teacher competencies-based PLE design and development framework, which contributes to the recommendation feature of the PLE concept. The Guided PLE Model aims at supporting today's learners successfully take control of and manage their learning. The Model achieves this aim with four objectives, which include: to identify the type of support required to build a useful PLE model, to determine teacher competencies for a PLE-based pedagogy, to build a recommender system and a PLE application on top of the Guided PLE Model, and, to find theoretical foundations of the Guided PLE Model and the GuidedLearn PLE application. This thesis uses mix-methods and multi-dimensional measure based design science research methodology to achieve these objectives. To this end, the design science research-based theoretical, as well as empirical grounding processes, are applied through conducting critical analysis. structured communication, theory analysis, and user experiment. Thus, to overcome the challenges that are brought about by information-overload, knowledge-deficit and personalisation-related issues, the Guided PLE Model provides present-day learners with semantic-rich user-user recommendations of people who match them on the basis of their skills. Therefore, as the proof of concept of the method, the Guided PLE Model contributes teacher competencies based and Latent Semantic Analysis driven recommender system called SkillsRec recommender (Skills-based Recommender), as well as a server-side PLE application called GuidedLearn PLE application. The SkillsRec recommender generates user-user recommendations in ranked order that it develops from the Semantic Analysis of teacher competencies for PLE-based pedagogy and learner interests. The GuidedLearn PLE application a front-end application platform of the SkillsRec recommender. is a prototype PLE application that follows bottom-up socialmedia. systematic and Human-computer Interaction design principles of the online learning environment designs. The applicability and the effectiveness of the Guided PLE Model have been demonstrated in this thesis through a four experiment-based usability study of the SkillsRec recommender. Which was conducted on 20 realprofiles of users over the GuidedLeam PLE application. Thus. SkilisRec recommender was compared with the ConIR recommender (Conventional Information Retrieval based recommender). The context of the four experiments was strongly tied to the classroom-based collaborative work and learning. This thesis discusses the results of the user evaluation study 0 t- two recommenders and provides study conclusions and implications. Thus, for target learner, the significant differences in user-skill similarity scores and in the number of skills returned by the two recommenders (i.e. 28 versus 19) at as high cut-off as 0.9 confirmed that the SkilisRec recommender can perform better than the ConIR recommender. The moderate values of precision and recall tests against the four experiments suggested that the SkilisRec recommender can yield more relevant user-skill similarities and user-user recommendations. Through generating recommendations that are developed from users' profile statements and not from users' interaction with the system or search history data, the SkilisRec recommender provides a simplistic solution to the Cold-start problem. the inherent problem of Collaborative Filtering based recommender systems. Thus, considering as base case the unique skill-similarity based user-user recommendation approach of the SkillsRec recommender that is developed on top of the Guided PLE Model, it can be concluded that the Collaborative Filtering recommendation approaches in particular and the learning systems which rely on learners' interests and skills will benefit greatly.

The full text of this document is only accessible to authorized users.