Abstract
In various disciplines, hierarchical clustering has been an effective tool for data analysis. However, traditional hierarchical clustering algorithms are not scalable to very large databases because of their high computational cost. To partially circumvent this drawback, in this paper, we propose a new algorithm for hierarchical linkage clustering as a solution for an efficient as well as reliable data clustering problem. Basically, our algorithm consists of two stages. In the first stage, the traditional linkage algorithms are applied to cluster a size-reduced version of the original dataset via boundary point detection. In the second stage, a k-nearest neighbors based classifier is employed to assign a cluster label to the remaining data points. Finally, evaluation is performed to show that the proposed algorithms can obtain good results not only in terms of the consumption of reasonable run times but also with better accuracy.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have