exploiting learner models using data mining for e-learning a rule based approach

A. Peña-Ayala (Ed.): Intelligent and Adaptive ELS, SIST 17, pp. 77–105. © Springer-Verlag Berlin Heidelberg 2013
78
M. Holzhüter, D. Frosch-Wilke, and U. Klein
Chapter 4
Exploiting Learner Models Using Data Mining for E-Learning: A Rule Based Approach
Marianne Holzhüter1, Dirk Frosch-Wilke1, and Ulrike Klein2
1
University of Applied Sciences, Sokratesplatz 2, 24149 Kiel, Germany {marianne.holzhueter,dirk.frosch-wilke}@fh-kiel.de 2 University of Kiel, Zentrum für Geoinformatione, Boschstraße 1, 24118 Kiel, Germany uklein@gis.uni-kiel.de
Abstract. The need for innovative didactical methods in combination with the efficient deployment of technical systems is an increasing challenge in the research field of e-learning. Research activities concerned with this observation have led to the understanding that the concept of learner models offers a range of possibilities to develop optimized, adaptive e-learning units (e.g. Graf et al. 2009, O’Connor 1998). Process information can enhance these approaches. Data mining is able to build process models from event logs. It means that formation about real process execution can be deduced by extracting information from event logs rather than by assuming a behavior model which has been built by conventional modeling methods. This applies to the e-learning context, because a certain behavior of an underlying process model tracked in a Learning Management System (LMS) may differ from the one assumed by instructors or learning object designers of elearning units. Instructors who need to attribute certain tasks to a huge group of online learners may not be capable of monitoring all factors influencing the appropriateness of all learner-task associations. Learning paths in LMS to which instructors have not paid attention to yet are of considerable interest. We apply a concept of rule based control of e-learning processes based on the framework we have presented in Holzhüter et al. 2010 to demonstrate these goals.
4.1 Introduction
Reflecting learner activities (especially to improve learning efficiency) in LMS gains importance as computer support of educational processes increases. The choice and deployment of LMS is associated with risks. Especially if personalization is necessary, high implementation and maintenance efforts (Dagger 2006) and costs (e.g. Kleimann 2008) can be opposed to lack of user acceptance. Lack of acceptance is often caused by the absence of plausible arguments which convey the advantages for learners, instructors and LMS
4.2 Related Work
As this research project partly is assigned to the geographical department of the University of Kiel, we especially focus on educational technology issues in the geographical field. Klein 2008 conducted a study which suggests that the discrepancy between employed media and the pupil’s interest (in combination with assumed learning effectivity through the media of interest) reduces the quality of geographical teaching. These two phenomena (discrepancy between media and interest and teaching quality issues) further were observed to reinforce each other. Especially schools (the target group of Klein’s study) often lack means to employ media which pupils consider interesting and helpful for learning. E-learning help by supplying simulations, interactive maps or playful test trainings in a structured and efficient way of material presentation is desirable. This is where process mining in combination
platform operators (see e.g. Tynjälä and Häkkinen 2005 for an illustration of the considerable range of challenges concerning e-learning projects). The improvement of e-learning process efficiency provides sound arguments to perceive the benefit potentials for learning and teaching scenarios. The knowledge discovery discipline of data mining has developed to the extent that the extraction of useful information can be applied to several kinds of systems – including e-learning systems (Frias-Martinez 2006, Tynjälä and Häkkinen 2005). Data mining can efficiently support the generation of learner models. They support the adaptation of LMS to individual learner needs (Nguyen and Phung 2008). A particular type of LMS is explicitly dedicated this approach: Adaptive Educational Hypermedia Systems (AEHS). The combination of these methodological and technological advances with findings from Business Process Management entails considerable potentials: It helps to focus on a learning process perspective exceeding the mainly isolated view on datasets which dominates in traditional data mining. Process mining, a specific sub-discipline of data mining, offers a useful set of concepts and tools to follow this learning process optimization and adaptation approach. The chapter is structured by seven sections, partly further structured into subsections. The two next sections provide the terminological and theoretical foundations need to be laid. Sect. 4.4 is divided into two subsections: The first deals with our approach to combine the process mining and the learning style approach – as a method of learner modeling. The second subsection introduces the overall integration approach of rule based control in e-learning. Sect. 4.5 is also split into two subsections: The firstly presents an architecture concept for our approach, ideally based on already existing technology in an institution (subSect. 4.5.1) and the second one deals with a possible implementation scenario based on open source software (subSect. 4.5.2). The second approach is useful for institutions that are in the process of building up a completely new architecture and do not dispose of experience regarding the available technologies.