An agglomerative technique for Pearson Type II mixture decomposition with applications

Mixture modeling can be considered as the probabilistic counterpart of fuzzy clustering. Gaussian mixtures are the most widely used distributions in mixture modeling and the expectation maximization (EM) algorithm is commonly used for Gaussian mixture decomposition. Gaussian mixtures are not suitable for certain problems. Moreover, the EM algorithm suffers from the disadvantage that the number of components in the mixture needs to be specified In this paper we introduce Pearson Type II mixtures, and present an agglomerative technique for Pearson Type II mixture decomposition which automatically determines the number of components required to model the data efficiently. We apply the proposed algorithm to detect lines and planes, and to classify 5 benchmark data sets. The results obtained are compared with results from a well known fuzzy clustering technique, the algorithm of Gustafson and Kessel (1979), and with agglomerative Gaussian mixture decomposition.