Applying Clustering Approach in Blog Recommendation
Zeinab Borhanifard, Behrouz Minaei Bidgoli, Hamid Alinejad‐Rokny · Journal of Emerging Technologies in Web Intelligence · 2013
The web has met a significant growth in using weblogs during the recent years. According to the large amount of information in the weblogs, bloggers are facing difficulties to find blogs with similar thoughts and orientations and their popular information. While there is a vast overload of information for blogs, it necessitates having a blog recommender system. Collaborative filtering is a well-known technique in recommender systems. This technique extracts the relations between users and items in according to its neighbor's ratings, and since users have rated just a small part of data, sparsity makes problems for collaborative filtering. This problem leads to an inaccurate comparison among users, and consequently it decreases the accuracy of collaborative filtering algorithms. The use of clustering technique decreases data sparsity and it improves system scalability. We have used clustering to recommend the blog while the blog have reciprocal role, and each blog is both as a user and as an item in the network. In this paper, we use graph clustering based on users' information about social network and we propose blog recommendation framework to get recommendations. Experiments on ParsiBlog 1 data indicated that application of clustering technique with collaborative filtering is better performed that traditional collaborative filtering algorithms, PageRank and etc. A comparison between PageRank algorithm and clustering application showed that graph clustering in recommender system could makes better results in terms of accuracy, quickness and scalability.