Year

2018

Degree Name

Doctor of Philosophy

Department

School of Computing and Information Technology

Abstract

1 Abstract Mobile learning through open educational resources (OERs) evidently differs from its traditional ways as it becomes more fragmented. Hence, micro open learning, where learning occurs within relatively short and fragmented time spans, has gained high popularity in the era of mobile and pervasive computing. However, adaptive micro learning support remains inadequate by current OER platforms. To address this, this thesis has leveraged the advantages from cloud computing and big data techniques to promote adaptive micro open learning. A service-oriented personalized learning environment (PLE) is realized by a newly designed Software as a Service (SaaS), namely Micro Learning as a Service (MLaaS). This smart system aims to deliver personalized OER with micro learning to satisfy learners’ personal demands in real time. It customizes adaptive micro learning contents as well as provides learning path identifications tailored for each individual learner. It consists of an offline computation and an online computation domain to provide recommendations jointly to improve computation performance and respond in the granularity of seconds.

To personalize the micro learning, this thesis has profiled the learner and learning context in an extensible manner with respect to the learning environment in space (i.e. mobile) and time (i.e. micro) manners. A dynamic learner model is conceptually built with regards to the internal and external factors that can affect learning experience and outcomes, while the internal factors contain personal intellectual and non-intellectual factors. A mechanism of the categorization and customization of OERs is also developed in the micro learning context, along with the measurement of micro OERs. This thesis has constructed a knowledge base to support the decision-making process of MLaaS. The knowledge base is built using a top-down approach. A conceptual graph based ontology construction is first developed. An educational data mining and learning analytics strategy is then proposed for the data level, acting as the main method to understand learners’ behaviours and recognize learning resource features.

However, this novel delivery and access mode of OERs suffers from the cold start problem because of the shortage of already-known learner information versus the continuously released new micro OERs. The learning resource adaptation still requires learners’ historical information. Also, relying solely on the offline computation, the existing recommendation may not guarantee reasonability and timeliness. Hence, an online computation solution has been introduced to assist OER providers and instructors dealing with the sparsity of data in OER recommendation. To compensate the absence of this information initially, a predictive ontology-based mechanism has been set up. A lightweight version of learner-micro OER model works in accordance with classification algorithms, spreading activation, and demographic similarity based inference to predict fresh learners’ features without the requirement of any known priori probability. An algorithmic framework is provided to realize the novel OER recommendation system based on heuristic rules.

These rules can also optimize the approaches to blending new-coming micro OERs into the established learning paths. Therefore, as the first resource is delivered to the beginning of a learner’s learning journey, the subsequent micro OER recommendations are also optimized using a tailored heuristic. Comparing with different widely used recommender systems, our evaluation shows that the proposed heuristic algorithms for online computation perform satisfactorily in terms of precision and recall values.

Share

COinS