The Research Of Personalized Service On WWW Based On Web Data Mining

With the rapid development of Internet, the number of web pages on web servers is growing exponentially. Web has already become a necessary and important way by which people obtain various information. However with the much of web pages providing abundant information resources to people, it also bring a new challenge that how to acquire information needed from web quickly and accurately. At present, the service method adopted by web is passive and single. Different users on the same website obtain the same information, but the need of users is very different. The existing information service system mainly exists the following limitation: browsers maybe lost themselves in the complicated information space because of existence of the big quantity of web pages and the hyperlinks between them. This is called "get lost in information". The great number of hyperlink which provided by web pages are useless for particular user and the useful content is concealed in numerous useless links. The interface that the information servers provides usually single and boring and it take on the same face to all. Current information service are still a kind of passive information providing way to server. Tt can't feel a customer's need actively, and can't provide active information service to user.The personalized information service can learn interest and behavior of user by collecting and analysing user's information, and realize the purpose of recommend information actively. It can change web service type from website center to customer center, and from one to many to one by one. So it can provides service to user high efficiency. The personalized information service technology of how to provide accurate and proper information aimed at user's personal characteristic has become a research hotspot.This thesis focuses on above needs and the shortages existing in current personalized service, and puts forward a Personalized Active Information Service System Based on Genetic Algorithms Transaction Cluster (for short GA-PAIS ).This system clusters transactions through genetic algorithms , and overcoming the shortages that exist in personal recommend algorithms: the complicacy is too high and sensitive to noise data etc. Experiments show that the algorithm is efficient on transaction clustering. For calculation the web page recommendation coefficient, the match factor and distance factor which would influence recommend accuracy are considered comprehensively. This thesis puts forward a calculation method of match factor based on Double Windows. The problem existing in current recommend algorithm, such as single recommend factor and neglecting page sequence will be resolved efficiently.