Mining educational data to provide intelligent information and personalized recommendations

点击翻译按钮获取中文摘要

Online courses have become very popular in recent years. While such courses have many advantages, they also have drawbacks, including high dropout rates, little or no feedback for educators on how learners learn, and little or no personalization for learners. Other fields such as business have successfully used data analytics, data mining, user modelling and personalization to address similar problems but educational systems currently do not employ such techniques. The primary objective of this research program is to advance research in data analytics, data mining, user modelling, and personalization in the educational domain through designing and evaluating algorithms and techniques to analyze and mine educational data, create user profiles, and use these profiles to provide users (learners and educators) with personalized information and personalized recommendations for improvement. This research program aims to create and evaluate educational data mining algorithms and learning analytics techniques that can not only identify effective and ineffective behaviour patterns of learners and educators but also identify learners who are at risk of dropping out or failing a course. These algorithms and techniques will then be enriched with approaches from the fields of personalization and user modelling to advance current research and create truly personal profiles of learners and educators. At Athabasca University, I have access to data gathered from over 850 online courses - a dataset that is majestic in size and that can be used to produce reliable and generalizable results. Furthermore, this program aims to develop and evaluate artificial intelligence and recommender system techniques that will use the information generated about learners' risk levels as well as effective and ineffective behaviour patterns of learners and educators to provide personalized recommendations. Personalization will be applied on various levels, and an automatic feedback mechanism will be developed to further enhance the accuracy and usefulness of the recommendations. The final result will be an open-source tool suite consisting of a set of software solutions for learners and educators that can not only be integrated with common online learning systems but also used and built upon by other researchers. These software solutions will act as personal coaches for the learners and educators, giving them information about their learning and teaching processes (e.g., learner risk levels and the effectiveness of learner and educator behaviour patterns). These coaches will also provide personalized recommendations on how to change behaviour patterns to (a) improve the effectiveness of teaching and learning, (b) reduce dropout rates, and (c) improve outcomes for the learners. In addition, this program will demonstrate the potential of using data mining, data analytics, user modelling, and personalization to improve online learning systems and to benefit education across Canada.