The Research And Implementation Of Personalized Recommendation Considering Web Content And Web Structure

Web personalized recommender systems anticipate the needs of web users and provide them with recommendations according to their navigation patterns. Nowadays, the research of personalized recommender systems has attracted a lot of attention. Such systems have been expected to have a bright future, especially in e-commerce and e-learning environments. However, although they have been intensively explored in the web personalized recommendation fields, and there have been some commercialized systems gradually, the quality of the recommendation and the user satisfaction of such systems are still not optimal.This paper designs and implements a novel web recommender system, which combines usage data, content data and structure data in a web site to improve the quality of web site recommendation. The contents of this paper are as follows:1. Collect web data from server log, page content and site topology as data source of the web personalized recommender system, furthermore, make preprocess according to the characteristic of Chinese web pages and recommender system, for the purpose of acquiring users'navigational patterns more exactly.2. Analyze the drawbacks of traditional transaction identification methods, and propose an improved one, which combines content data of web pages, and applies document clustering algorithm in this process. In addition, modify the clustering algorithm according to the need of the proposed method.3. Cluster these content coherent transactions to generate primary users'navigational patterns. Furthermore, augment these patterns with their linked neighborhoods, and compute weight based on site connectivity, by which these pages in recommendation list are ordered.4. Recognize the current focused topic of interests to the active user as his navigational pattern. Define the page recommendation ranking function, considering that whether a page has been visited by active user in his current session should be distinguished among pages to be suggested. Design and implement a recommendation algorithm with high accuracy and low time...