LF-CARS: A Loose Fragment-Based Consensus Clustering Algorithm with a Robust Similarity

摘要

The consensus clustering technique is to combine multiple clustering results without accessing the original data. It can be used to obtain the clustering result from multiple data sources or to improve the robustness of clustering result. In this paper, we propose a novel definition of the similarity between points and clusters to represent how a point should join or leave a cluster clearly. With this definition of similarity, we desigh an iterative process which can determine the number of clusters automatically. In addition, we propose the concept loose fragment which is improved from clustering fragment into our method for speed-up. The experimental results show that our algorithm achieves good performances on both artificial data and real data.