Subtractive Clustering based Distributed Gaussian Mixture Model for Density Estimation and Clustering in Sensor Networks

This paper presents a subtractive clustering-based distributed Gaussian mixture model (GMM) in sensor networks. In literature, the Expectation-Maximization (EM) algorithm is frequently used to estimate a mixture's parameters. An inaccurate estimation would deteriorate the extracted data from the obtained model. Once we estimate the optimal number of components, two problems would have been solved: estimation of the optimal number of clusters and estimation of the initial values of the model parameters. In this paper, we propose a distributed data mining algorithm for sensor networks. The proposed distributed GMM is based on the subtractive clustering algorithm which is noise robust method. To estimate the optimal number of clusters, our proposed algorithm uses the mutual relationship between the mixture components. Experimental results show the effectiveness of the proposed method which estimates both the optimal number of clusters and initial mean vectors.