Activity Sequence Modeling and Multi-Targeted Clustering for Personalization in E-Learning

Research output: Types of ThesesDoctoral Thesis


The impact of e-learning on the global learning process has been steadily growing over the past decades. Not only has e-learning, in different manifestations, made its way into the majority of educational settings, it has in many aspects caused a revolution of learning environments and a paradigm shift regarding the general understanding of learning itself. While the traditional focus of e-learning lay on the individual learner, the theories and practices of learning collaboration of the physical world were progressively transferred into electronic counterparts. At the same time, the role of the learner changed from the one of a passive consumer of content to the one of an active participant, and learning in general shifted from a teacher-centered process to a learner-centered one. In parallel to the change of the focus, also the technology used to support learning evolved. The general area of information systems underwent a process of steady growth in several directions, resulting in the information pool becoming insurmountably large for any single person to perceive, analyze and comprehend. This development gave rise to the need for information filtering and selection facilities separating relevant from irrelevant pieces of information. As learners are different in various ways, including their interests, background, knowledge, long-term and short-term goals, their motives for interacting with the system, etc., there is no way of universally assessing the relevance of information, leading to the need for personalization. In the area of e-learning, adaptive software systems autonomously tailor the appearance and amount of learning content, learning paths, and learning support to the individual learner's or groups of learners' requirements and characteristics. This thesis focuses on the analysis and interpretation of learner activity data in order to provide a basis for reliable conclusions regarding learners' individual needs and requirements. It proposes a general approach to the utilization of activity data in the learner modeling process, emphasizing the fact that learner activities can not necessarily be treated as independent from each other, but might be interrelated. In more detail, an activity sequence modeling method is presented and discussed in combination with an unsupervised learning approach applicable at different levels in order to: detect predefined, well-established problem-solving styles in students' problem-solving sequences; discover new problem-solving styles along predefined learning dimensions; and discover potentially interesting learning dimensions and associated problem-solving styles. Deliberations on how the gained pieces of information about learners' problem-solving behaviour can be fed back into the process by offering individual adaptations based on them, complete the cycle of adaptation. Furthermore, adaptivity in the area of e-learning in general, and in relation to the proposed approach, is discussed from the perspective of security and privacy, issues that are of high relevance in systems that rely on the collection and interpretation of users' (personal) data.
Original languageEnglish
Publication statusPublished - 2011


  • Adaptive Systems
  • Personalization
  • Activity Sequence Analysis
  • User Modeling
  • Clustering
  • Machine Learning
  • E-Learning
  • Activity Sequence Modeling
  • Learning Behaviour Analysis
  • Problem Solving Styles


Dive into the research topics of 'Activity Sequence Modeling and Multi-Targeted Clustering for Personalization in E-Learning'. Together they form a unique fingerprint.

Cite this