Applying Data Ming To Recommendation Systems

Data Mining(DM) is a fast developing research area in the last decade. It integrates theories and technologies from Database, Artificial Intelligence, Mechine Learning and Statistics, and has become a bridge between theory study and real world applications. For it is a cross-fields subject and can be used widely, DM has become one of the most active research areas for researchers and business units. Recently Web Mining is a focus of DM. All kinds of data and new application needs come from INTERNET bring big chanllenges for DM. Web Intelligence (WI) is a new research area exploiting Artificial Intelligence and advanced Information Technology on the Web and Internet. Data Mining plays an important role in Web Intelligence. Web-based intelligent system is an important aspect of DM and WI. Intelligent Recommendation System is a kind of typical web-based intelligent application system. With the rapid growth of the World Wide Web, people more and more need a method to help them selecting useful information from everyday's huge data, which brings big demands for intelligent recommendation systems. My thesis focuses on Bayesian Network (BN) learning and inference. It presents a Bayesian Network learning algorithm based on Conditional Independence method and an inference algorithm based on random probability. It gives analyses of the Bayesian Network learning and inference algorithm. Take Bayesian Network as a model, this thesis designs and realizes a web-based intelligent recommendation system---PIR (Personalized Intelligent Recommendation) system. And then expands it to an open framework for recommendation system. Based on PIR framework, we can integrate new algorithms easily. The recommendation component can...