Modifications for fuzzy C means clustering to enforce ellipse and parabola shapes on clusters

Abstract

The work presented in this Thesis is mainly focus in the area of clustering, to be more specific in the area of Fuzzy C Means Clustering. The problem is to change the search geometry of the above mentioned algorithm. The shapes which the modified algorithms are required to enforced are Ellipse and Parabola. The task related to shapes were done in [10] and [14]. We try to develop a simpler approach. The cost function was altered and update equations were derived. The algorithm was also modified and tested on artificial datasets. One of the algorithms is also implemented in image processing problem.