Process Models Enhancement with Trace Clustering

Abstract

Learning Management Systems collect data (such as event logs) about learners and trainers. There are techniques for analysing this data such as Educational Process Mining. In our previous work, we proposed an approach that extracts knowledge about learning paths based on process mining algorithms by generating process models. The latter are used for learning resource recommendation by taking into account learning features (learning style, interests, learning results, etc.). This approach is available for a limited size of event logs. In fact, process models generated from event logs of large classes are not expressive. Trace clustering is one of the successful methods that lead to overcome this limitation. For this reason, we aim to improve the previous approach by using trace clustering in order to characterise learners before the discovery of the corresponding process models. We applied the proposed approach on a Moodle dataset of 100 undergraduate students. Results show that trace clustering improve the general quality of discovered process models.