Based On Internet Users Personalized Interest Model

With rapid development and popularity of the Internet, almost every aspect of our work, study, life and entertainment have been changing gradually, and meantime many problems needed solved quickly are appearing. For example, how to resolve the problem of "information overload" and "resource lost", and help the people to make the best of the information and resource of Internet, by developing more intelligent and personalized Internet systems. In this thesis, personalized User Profile is discussed, which is the base of various intelligent and personalized systems. So the study has theoretic and practical value.Firstly, the paper puts forward the key techniques to buld Personalized User Profile, such as Web Mining, User Action Mining, Machine Learning, and Agent. Web Mining is developed by the combining of the traditional data mining and the new web techniques, and is to extract the interested potential useful pattern and hidden information from the web files and web activities. Three Web Mining techniques are put forword in this paper: log mining,content mining and struct mining.As another key technique, User Action Mining is also used in building personalized User Profile. In respect of Machine Learning, the paper discusses the neural network algorithm which is a well-studied method in the realm of Aritificial Intelligence. And agent as a new method widely studied and used by all kinds of intelligent Internet systems today, is also discussed in brief.Then the paper puts forward the architecture of the personalized user interest model. The algorithms of each part of the system are analyzed and designed, including assessing the user interest degree of a web page, cutting words from text, extracting feature words, building and updating user profile by hybrid mining and learning algorithms. Some experiment is also discussed. To obtain user interest, the first step is to analyzed the relativity between the pages read by user and the user inerest, which is named "User Interst Degree" and is analyzed by observing the user behaviour (such as bookmarking, saving,printing,browsing,and scrolling). At the same time, the ability and the degree of how different user actions to indicate the user interest are also discussed. The second step is to analyzed what is the real user interested content in the web page by combining the web mining and machine learning method, changing the html documents to Vector Space, extracting feature, and learning user interest by relativity learning method and neural network.Finally the paper discusses some applications of Personalized User Profile and designs a Personalized Active Information Service Model. Personalized Active Information Service Model is also named Personalized Information Recommendation System. The steps to build the system are building the user profile, searching and filtering the information intelligently, recommending the information actively and updating the user profile according to the user's feedback. The system can also help the people to share the information with users of Interest Group. Some other applications are also discussed, such as intelligent searching engine, adaptive web site and personalized study based on Internet.