Abstract:

Adaptive learning provides students with access to individually tailored tutoring, i.e. an automated way of learning that adapts to different personal, academic, social/emotional, and/or cognitive characteristics of the particular learner. There are two types of adaptation for personalisation in learning systems: person-driven and system-driven. The paper discusses system driven approach, concentrating on recommender systems and giving an insight into the use of data mining techniques on Web access logs for recommendation of pages that are most likely to be selected by the learner. History of access sequences and modern heuristic methods applied to access data mining are being used to evaluate and understand learners’ access patterns. Understanding navigation preferences of the learner can enhance the quality of the content recommended to him, and thus made learning more adaptive and attractive to particular student.

The information gathered through the web is evaluated using web mining: web content mining, web structure mining and web usage mining. Web usage mining philosophies as well as architecture for usage-based web personalisation are reviewed in the paper. Three main web usage mining phases (data pre-processing/preparation, pattern discovery and pattern analysis) are briefly described, highlighting main aspects of the process.

Web usage mining is not a trivial process, and there are several data mining algorithms and methods developed from statistics, machine learning, data mining and pattern recognition being used in web usage mining process. Which of them must be used depends on what kind of insight is needed. For example, for prediction of the next event sequence mining can be used, and for discovery of the associated events association rules applied. Each dynamic web usage mining technique has its own characteristics and can serve the implementation of particular personalisation purpose.

In the paper, first of all, systematic review on existing personalisation techniques is made, concentrating on recommender systems, presenting generalised model of recommender system and listing content filtering techniques used by recommenders. Second, literature review on web usage mining is made, explaining the nature of web usage mining process. Third, web usage mining techniques used by recommender systems are reviewed and briefly described. Finally, a research on how to overcome the challenge of large amount of provided recommendations is described, presenting recommendation dashboard solution. Recommendation dashboard enables to explore and manage received recommendations and helps to organise and highlight the best ones.

It is concluded that a synergy of recommender systems and recommendation dashboards provides useful support for recommendation process, enabling learner to successfully adapt learning content to his or her learning goals.

Keywords:

Web usage mining, recommendation dashboard, recommender systems, pattern discovery, path completion.