Web page classification and hierarchy adaptation
Brian D. Davison, Xiaoguang Qi · 2012
Classification is a supervised learning problem in which a classifier is trained on a set of data labeled with predefined categories and then applied to label future examples. It plays a fundamental role in a number of essential tasks in information retrieval and management. Advanced classification approaches will benefit systems that search or manage web information, as well as other types of information in general. In this dissertation, we investigate methods to improve classification from two aspects: feature enhancement and hierarchy adaptation. For feature enhancement, information from the neighboring pages on the web graph is studied. Novel methods to effectively utilize such neighboring information to improve classification are proposed and analyzed. For hierarchy adaptation, evolutionary computation methods are used to search for better hierarchies in order to improve hierarchical classification. We also investigate problems that impede user navigation in hierarchies, and propose novel methods to facilitate efficient navigation. Experiments on multiple real-world datasets show that the proposed approaches can significantly outperform previous state-of-the-art methods.