On c-Regression Models with Constraints on Cluster Size and Its Kernelization

Abstract

Clustering is a data mining tool that divides a dataset into some groups to obtain useful knowledge. The c-regression models is a clustering method to divide a dataset into linear shapes by assigning objects to linear regression models. K-member clustering is a clustering method with constraints on the total number of objects in a cluster to at least K or more. We have proposed COntrolled-sized c-Regression models (COCR), which combine these two methods and classify the dataset roughly even. Considering more practicability, the division of the nonlinear shape may be required. In this study, we propose a clustering algorithm that divide dataset into nonlinear shapes and roughly even.