Nearest neighbor method with priority selection of denser clusters by analysis of similarity matrix for radar pulse clustering

Clustering is used in radar pulse classification and de-interleaving. Proper selection of initial centers for clusters is the fundamental problem in clustering methods. In the proposed algorithm, the similarity matrix for the input data based on defined neighborhood radius is created. Then, with analysis of similarity matrix and selection of lines with the most similar code, denser clusters are separated respectively. In this way, regarding to the high stream radar pulse, without requirement to iteration, pulse strings are separated as optimal. This algorithm has capability of priority selection of dense categories in clustering. On the basis of neighborhood radius defined, data on the common border clusters, have been carefully separated.
In particular, the advantage of using this algorithm is selection of dense clusters center in radar pulse clustering. The proposed method can also be used to cluster the data in various fields. . Results of proposed method for data sample consisted of 200 radar pulse is compared with the results of clustering around the leader which is one of the main clustering algorithms in the field of radar pulses. Evaluation and measure of clustering validity such as Dunn, Silhouette and RMSSD indexes, shows efficiency of the proposed algorithm.