Modeling and Detecting Drift in User Interest Based on Hierarchical Classification

摘要

During participating in Web 2.0 applications, people enjoy the convenience brought by Internet resources, but also encounter the problem of lost in Internet. This paper takes the Douban movie review as the background, and proposed a method for detecting the drift of user's interests. The user's comment and the frequency with which the user watches a certain type of movie are used to determine the extent to which the user is interested in the movie, and the parameters p and q are used to adjust the degree to which the two factors affect the interest weight. A hierarchical user interest classification tree (named HC-tree) and time window table are designed to maintain the interests and their weights in multi-granularities way. Also two functions λ and ψ are designed to quickly identify the drift of user interests. A large number of experimental results show that the proposed method is better on the correctness for user interest drift detection and is prior to other similar algorithms.