A novel analysis on shell type synthetic data sets

In this study, a novel clustering analysis was performed on shell-type synthetic data using the K-means method according to specified parameters. Two-dimensional synthetic data sets were used because real data sets are usually multidimensional and therefore cannot be visualized. Since the data points in multidimensional datasets are mostly concentrated in the outer shell, the datasets were created in shell type. In image processing studies, the presence of data in the outer layers of images with edges removed can be given as an example of shell type data. With the analysis study, the most successful clustering performances of shell type data sets with different topologies created in two-dimensional space according to the parameters of inter-cluster spacing, intra-cluster spacing, number of clusters, shell thickness, number of data and number of nested clusters were observed with computing time and accuracy criteria and the optimum data set structures were tried to be found. In addition, the relationship between the statistical properties of the used KTSVs such as skewness and kurtosis and clustering performance is investigated. The results of the experiments are presented in tables and graphs.