Webpage Recommender System concerning high dimensional and sparse features

In this paper, we design a Webpage Recommender System which clusters users based on users´ browsing history to implement collaborative filtering and prepares webpages that may arouse their interest. In order to solve the problem of high-dimensionality and sparsity in collaborative filtering, the proposed system clusters users using CABOSFV, an efficient algorithm for high-dimensional sparse data clustering of binary attributes. And it uses an automatic webpage classifier to solve problems of cold-start and exhausting of recommendations. When online users send requests to the system, it responses them with those well prepared recommendations.