CLARANS of Uncertain Objects Based on Bayesian Probability
Tong He · International Conference on Electric Information and Control Engineering · 2012
In CLARANS clustering algorithm of uncertain objects each of which is given as a Gaussian mixture model, Kullback-Leibler divergence is used as distance measure of the algorithm. Meanwhile, in the past research, the effect of Bayesian probability is as good as Kullback-Leibler divergence when it is used to be distance measure between Gaussian mixture models. Sometimes, the former is even better than the later. So by using Bayesian probability as distance measure, a new CLARANS clustering algorithm of uncertain objects which is based on Bayesian probability is proposed in this paper. In an extensive experimental evaluation, we show different effectiveness and efficiency of two clustering algorithm which using different distance measure.