The Research On The Key Technology Of User Modeling Base-on URL

With the rapid development of the Internet, the types of network applications and network information are gradually increasing, and different users’ demand for information or services is different. In order to provide users with targeted information and services, personalized service came into being. User modeling is the key part of personalized service, so user modeling technology has important research value for understanding and characterizing of users, as well as providing users with more quality services.The web sites that user has visited have a very important role in modeling. Most of the traditional Web user modeling methods are based on the mining to the web page contents. These methods have some limitations because it is very time consuming to get web content, and it is difficult to excavate the information in the web page. But the user modeling based on URLs has the advantages of short time, simple processing and so on. In this paper, we try to establish a user interest model and a user attribute modle through in-depth analysis of URLs, in which the user interest model can describe the user’s interest preferences and user attribute model can describe user attribute information.In the process of user interest modeling, the frequency of a user’s access to the websites in one topic can reflect the user’s interest in the topic, so webpage topic classification plays an important role in determining the user’s interest. In this paper, we introduce the N-gram language model to classify the web pages just using the URL. Then, we proposed a user interest modeling method. This method can classify webpage into different topics using URL, and then get user interest information. Finally, the user interest model can be established by the user interest modeling algorithm.In order to obtain the user’s attribute information, this paper proposes a user attributes modeling method based on URLs. Through in-depth analysis of URLs, a series of rules for the acquisition of user attributes are constructed. Then we can mine the user attributes according to the set rules from the user’s access records, and update the user attribute model. Finally, a parallel scheme is proposed to improve the processing efficiency.Using the research method of this paper, a user modeling system based on URL is designed and implemented using the access log of the VPN users. Experiments show that the system can quickly mine the user’s interest and attribute information from a large number of URL, and then build a complete user model. At the same time, the system has strong stability and high performance.