AbstractThe explosive growth in online learning materials has generated an urgent need for new techniques and access mechanisms in order to reduce information overload. E-learning recommender systems have become very popular as a means of delivering learning materials. The main challenge is how to recommend quality learning materials to students based on their different backgrounds, knowledge, interests and needs.
This thesis presents an integrated approach to recommending online video materials through sentiment analysis and hybrid filtering algorithms being used to support the development of personalised e-learning recommender systems and to address the information overloading and cold start problems.
This thesis makes five major contributions. Firstly, it proposes a personalised e-learning recommender system framework. PREFERS combines the techniques of student profiling, knowledge estimation, recommendation approaches and classification. Secondly, it proposes a hybrid recommendation algorithm integrated with sentiment analysis to provide personalised educational online video learning materials to students based on students’ profiles and the students’ feedback and interactions with the system. Thirdly, it develops a student profiling method using data mining techniques in order to construct a student profile and a student knowledge level from multiple feedback. Fourthly, it develops a dynamic hierarchical classification method/algorithm based on students’ feedback and the textual features of the videos to classify online video learning material using machine learning and information retrieval techniques. Finally, it designs and implements a personalised e-learning recommender system based on the PREFERS framework. The prototype system has been tested and evaluated with real users. Results show that the system can effectively recommend relevant learning materials to students based on their needs.
|Date of Award||Jan 2019|
|Supervisor||Linda Yang (Supervisor) & Salem Aljareh (Supervisor)|