Enhancements in Accuracy and Computation Time of Spectral Clustering

The spectral clustering is the clustering technique which cluster similar and dissimilar elements according to the dataset elements. The various techniques have been proposed to cluster similar and dissimilar data using spectral clustering. Among various techniques affinity matrices, mean shift algorithm are used to create similarity graphs and normalized algorithm is applied to cluster data. It has been analyzed that due to single use of normalization, cluster quality is reduced as some of the data points remain un-clustered. To improve cluster quality of the algorithm, Shi & Malik is used as normalization algorithm which leads to improve cluster quality and reduction in processing time.