A Study On The Mechanism And Methods Of The Personalized Recommendation Based On Web Mining In E-Commerce

In today's highly competitive e-commerce environment, the personalized recommendation has emerged as a critical application which is essential to a Web site to retain visitors and turn casual browsers into potential customers. However, there are wide gaps between our country and other developed counties about the personalized recommendation, which is mainly caused by the scarcity of academic research. Under this background, this dissertation applies the theories and methods of Web mining to the personalized recommendation in e-commerce, and utilizes the theory of comprehensive information and the model of information moving process to systematically research the principle and methods of the personalized recommendation in e-commerce.First of all, the process of the personalized recommendation in e-commerce is discussed, and according to the process, the personalized recommendation model in e-commerce is analyzed from the perspective of exploitation and utilization of"click stream"information resource based on the theory of comprehensive information and the model of information moving process, then the personalized recommendation method system is also proposed.Secondly, the problem of the Web user's interest analysis and recommendation based on the syntactic method is discussed. The framework of the Web user's interest analysis and recommendation is described firstly, and the process of discoverying Web transations from the Web logs is analyzed, then the method of the Web user's interest analysis and recommendation based on the cluster of Web transactions is discussed in detail. Thirdly, the problem of discoverying and matching recommendation rules based on Web text mining is discussed. The model of discoverying and matching recommendation rules based on Web text mining is constructed firstly, and the process of describing Web text based on Vector Space Model is analyzed, then the method of discoverying and matching recommendation rules based on the cluster of Web feature items is discussed at length.Fourthly, the personalized recommendation method based on Web semantic knowledge is discussed, which belongs to the semantic method of the personalized recommendation in e-commerce. The framework of this method is proposed firstly, and the construction process of Web domain ontology is illustrated by the example of the dangdang Website. According to the Web domain ontology, aggregate semantic Web usage profiles and user active session are discovered, then the personalized recommendation based on Web domain ontology is analyzed detailedly based on semantic similarity measurements among complex objects.Finally, the problem of constructing the utility function of the Web user based on the pragmatic method is discussed. The method of constructing the utility function based on the user implicit feedback is discussed firstly, and then a common user utility function of a special commodity (such as personal computer) is assumed. According to this function, a prior Bayesian network is established, and then the method of utilizing the learning mechanisms of Bayesian networks to construct the utility function is analyzed.