Mining User Intension with Fuzzy Theory and Clustering Technique for Learning Object Content Recommendation of e-Learning Systems

• Main Authors: Ruei-Yuan Guo, 郭睿原
• Other Authors: Cha-Hwa Lin
• Format: Others
• Language: en_US
• Published: 2006
• Online Access: http://ndltd.ncl.edu.tw/handle/74555375016722723206

碩士 === 國立中山大學 === 資訊工程學系研究所 === 94 === The perception of incorporating digital information into online educational systems and the ideal of developing digital schools for lifelong learning have drawn much attention of the governments, academia, and industries around the world. The techniques of interactive learning have become a primary research topic in E-learning. However, most existing E-learning systems provide static instructional materials. The techniques of dynamic learning content management that adaptive to individual user knowledge level and learning goals have been tough challenges for the related research communities. The resulting repetitive and blind learning phenomena have significantly reduced user performance and motivation. We hypothesized that new algorithms of adaptive learning based on the integration of current information technologies, the use of fuzzy theory to express the uncertainty features of the user knowledge, and the exploitation of clustering techniques to analyze the knowledge of a user for the comprehended areas of the domain knowledge will effectively improve user satisfaction. In this study, a prototype system is developed, implemented, and experimented by using SCORM run-time environment. The knowledge of teaching domain and the features of the learner behavior are modeled by ontology to represent the hierarchy and relationship of the learning concepts. To quantify user knowledge and learning ability, fuzzy sets are applied with multiple analysis dimensions based on the pedagogical strategies and user learning experiences. The performance of a user for learning knowledge concepts is then evaluated. In particular, an algorithm is designed to extract the existing learning paths of a user by the relative position of the concepts that the user attains in the domain knowledge. Furthermore, the candidate direction types for recommending concepts are inferred and the candidate learning concepts that are appropriate or inappropriate to learn followed up can be identified by rules. Moreover, the candidate learning concepts are scheduled to construct customizable learning routes by clustering techniques. The personalized learning contents that best matching user learning intention would then be presented to the user. Simulations study in the uniform and normal distributions for the grades of users is conducted to evaluate the tutoring model for three levels of users. The experimental results show that the proposed model helps different levels of users to learn the domain knowledge effectively and the accuracy of recommending the relevant learning object contents is superior than the random selection method. With a richer description of user knowledge and features, the proposed adaptive system for online learning assistance may better diagnose the understanding of a learner and enhance the pertinence of the retrieved courses to user intended learning to improve the service quality for the user.