Session F4E 978-1-4244-1970-8/08/$25.00 ©2008 IEEE October 22 – 25, 2008, Saratoga Springs, NY 38 th ASEE/IEEE Frontiers in Education Conference F4E-15 Forming Communities in Web-based Educational Systems through Users´ Preferences and Interest Measuring Reginaldo A. Gotardo, Cesar A. C. Teixeira, Sergio D. Zorzo reggotardo@gmail.com, cesar@dc.ufscar.br, zorzo@dc.ufscar.br Abstract - Service customizing in Web-based Educational Systems aims at directing content and teaching strategies towards students´ individual and group needs. This paper presents a virtual learning community forming approach, which helps knowledge exchange among its members. The approach here presented uses implicit and explicit information collection of users´ interests and preferences and relates the values in a correlation among the users. The correlation measuring ends up in groups with distinct characteristics. The proposal presented is validated by case study and it shows that from correlation values among the users in the interest and preference items, a group algorithm results in the formation of the intended groups. The results among interest and preference correlation were compared and assessed. Index Terms – Personalization, Recommendation Systems, Web-based Educational Systems. 1 - INTRODUCTION As the internet becomes more widespread, Web based Educational Systems (WbE-S) demand more and more attention. Another very important area on computing deals with Web systems personalization. The personalization applied to WbE-S is the focus of several works in research fields like Data Mining, Web Mining, User Modeling, Adaptive Hypermedia, Intelligent Tutoring Systems and Recommendation Systems [1]. The virtual communities or groups for learning used as a tool to content dissemination and to improve learning process is an important theme to distance education and content personalization [2]. The user modeling, or student modeling, has been one of the major areas of studies and challenges for Web based Adaptive Educational Systems. To model and adapt the user profile, information about the user’s behavior is needed. This was implicitly observed or explicitly asked to the user. In this work, we present the use of Collaborative Filtering techniques, a subset of Recommendation Systems, to forming student communities to the WbE-S. The selection of information for the Collaborative Filtering was performed in a traditional way: through the explicit classification of preferences in the content by the user in a WbE-S. However, this classification isn’t sufficient to well-formed interest groups and we used implicit information about user’s navigation, also called usage mining. Commonly, Collaborative Filtering uses explicit evaluation of the users about the system’s resources, and it doesn’t considerer the implicit relationship about them. The implicit information collected measure the interaction of the user with the system resources. The information bellows are divided in three types: Access Total Time, Most Recently Used and Most Frequently Used. However, these information are weighted by a tutor or specialist that knows the domain, in order to measure the user choice by his real behavior in the system. Through of measures of three user interaction values with the system, it is presented a different approach to obtain interests groups in a WbE System. This work is organized as it follows: section 2 presents the traditional Recommender Systems Approach, Web-based Education uses and Forming Communities information; section 3 presents how transform web usage in interest measures; the section 4 presents how preferences was measured; the section 5 presents our Approach to Forming Communities using User’s Interest and Preferences; section 6 discusses the case study results and section 5 presents the conclusions and future works. 2 - PERSONALIZATION, RECOMMENDATION SYSTEMS, WEB-BASED EDUCATION AND FORMING COMMUNITIES Personalization is an important issue in e-learning as it might help to improve both student performance and use experience. The common e-learning systems usually don’t allow the courses personalization to the student profile, but only propose standard courses. They are limited to direct appropriate learning to the student and don’t consider a student learning model. [3] E-learning systems, which uses user modeling and personalization, can allow the user receive more interest information and more adequate content to learn. There are several methods for e-learning personalization as Intelligent Tutoring System, Adaptive Hypermedia and the Recommendation System - method traditionally used in e-commerce applications. A Recommender System is a new approach to help users find relevant information. Basically, all the Recommender Systems do the same: they try to identify the most important items to the users and then recommend these items [4].