On Initialization of the Expectation maximization Clustering Algorithm
Iterative clustering algorithms commonly do not lead to optimal cluster solutions. Partitions that are generated by these algorithms are known to be sensitive to the initial partitions that are fed as an input parameter. A “good” selection of initial partitions is an essential clustering problem. In this paper we introduce a new method for constructing the initial partitions set to be used by the Expectation-Maximization clustering algorithm (EM algorithm). Our approach follows ideas from the Cross- Entropy method. We use a sample clustering provided by means of the EM algorithm as an alternative for the simulation phase of the Cross-Entropy method. Experimental results reflect a good performance with respect to the offered method.