Temporal Graph Partioning for Clustering in Tagging Systems

Today, information growth in the world of Web 2.0, due to the vast amount of data and change of some concepts over time, there is a lot of unnecessary and irrelevant information to what users are looking for. In this paper, we for solve this problem, propose temporal clustering of tags for systems that use tags as a metadata and are changing over time. The way we use for clustering, is temporal graph partitioning tags by changing the tag similarity weights during the time, then clustering will change and adapt itself with the changes. To demonstrate the effectiveness of this approach, we implemented it on a data set of MetaFilter site and compared it with similar methods. The results show that our proposed methods improved F-Measure out 24% compared to best clusters in the same way, over time, has improved and its concept is associated both with the past concepts and the newsletter.