Gaussian Hierarchical Bayesian Clustering Algorithm
Rafael Eduardo Ruviaro Christ, Edwin Villanueva, Carlos Dias Maciel · Seventh International Conference on Intelligent Systems Design and Applications (ISDA 2007) · 2007
This paper presents the Gaussian hierarchical Bayesian clustering algorithm (GHBC). A new method for agglomerative hierarchical clustering derived from the HBC algorithm. GHBC has several advantages over traditional agglomerative algorithms. (1) It reduces the limitations due time and memory complexity. (2) It uses a Bayesian posterior probability criterion to decide on merging clusters (modeling clusters as Gaussian distributions) rather than ad-hoc distance metrics. (3) It automatically finds the partition that most closely matches the data using Bayesian information criterion (BIC). Finally, experimental results on synthetic and real data show that GHBC can cluster data as the best classical agglomerative and partitional algorithms.