Automatic Personalization in E-Learning Based on Recommendation Systems: An Overview

Automatic Personalization in E-Learning Based on Recommendation Systems: An Overview

Mohamed Koutheaïr Khribi (University of Tunis, Tunisia), Mohamed Jemni (University of Tunis, Tunisia) and Olfa Nasraoui (University of Louisville, USA)
DOI: 10.4018/978-1-60960-842-2.ch002
OnDemand PDF Download:
No Current Special Offers


Web based learning environments are being increasingly used at a large scale in the education area. This situation has brought a dramatic growth in the amount of educational resources and services incorporated continuously in these systems, and related access and usage of this educational content by a diversity of learners. However, the delivery of this educational content is generally done in the same way for all learners without giving any special attention to the different consumption styles or differences between their profiles and individual needs. Therefore, providing personalization in e-learning systems has to be considered as a necessity and not an option. Recommending suitable links represents an instance of adaptive navigation support technology. E-learning recommender systems are used to locate relevant educational Web objects that better match the learner’s profile and interests, this requires the ability of a system to predict learner’s needs and preferences. Therefore, recommendation systems need to use Web mining techniques in one or more phases of the recommendation process, especially in the modelling and pattern discovery phase. Most emergent recommendation systems in e-learning tend to rely on automated detection of student’s preferences and needs since it is more efficient and attractive to provide needed support to students without requesting any explicit information from them. In this chapter, we present an overview of personalization in e-learning based on recommendation systems and Web mining techniques.
Chapter Preview

Automatic Personalization

Three main approaches have been used in automatic personalization, namely content-based filtering, collaborative filtering, and rule-based filtering techniques. These different techniques are distinguished by the type of input data collected to build user profiles, by the strategies used to build these profiles, and by the method used to make predictions or generally to provide personalized content.

Sources of the Collected data include (Eirinaki et al., 2003):

  • Content data referring to all web objects (pages, items, text, images, etc.)

  • Structure data which concerns hyperlinks connecting web pages to each other

  • Usage data representing users’ log history as a set of records containing web pages visits, time of visit and many other attributes. Such data can be collected from web server logs, cookies, system logs, several session tracking tools, etc.

  • User profile data which consists of a set of attributes describing a user such as demographic information and preferences. This information can be collected either explicitly (via questionnaires, forms, etc.) or implicitly (web server logs and system logs).

It is worth noting here that learning from collected data can be done online (Memory based learning) while the system is accomplishing the personalization task or offline before the personalization phase (Model based learning). Memory based systems tend to simply store all data and later utilize it at the time of providing recommendations, at a great online computational cost. In contrast, Model based systems avoid this online cost by performing the expensive computations in an offline mode and later using a learned model at the moment of recommendation at lower cost (Mobasher, 2007).

Once the collection of web data has been done, it should be preprocessed in order to prepare it in a suitable format compatible with the chosen analysis technique used in the next phase. Offline modeling techniques are typically based on Web Usage Mining which is applied to discover interesting usage patterns and relations between pages and users (Nasraoui, 2005).

Complete Chapter List

Search this Book: