A Classification Framework and Assessment Model for Automated Web Site Evaluation

This paper describes a neurofuzzy-based approach to automatically classify publicly accessible World Wide Web sites. The suggested methodology appears most relevant investigating corporate Web sites for advertising and customer support purposes but is equally valuable for analysing the hypertext structures of educational or non-profit organisations. Future efforts will focus on extending the framework from preand after-sales functions to customizable electronic transactions. A taxonomy is proposed based on "hard" criteria for classifying Web sites. Research methods and phases for a validated model are outlined and the most relevant attributes of such a system are defined. The described Web site analysis tool supports the automated data gathering and assures the necessary "critical mass" of training cases. The gathered data and the Web site taxonomy are fed into a neurofuzzy system which uses supervised learning in the assessment tool and unsupervised learning (fuzzy clustering) for validation.