A modified penalty function in fuzzy clustering algorithm

Clustering is one of the most challenging tasks to organize and categorize data with the presence of noise and outliers. With the increase of overlapping especially in gene expression data, clustering ability decreases considerably. A number of exigent algorithms are available to confront such data in the field of biomedical engineering. The proposed approach demonstrates the modified penalty function using co-variance of membership in the objective function of standard fuzzy clustering algorithm. This may resolve the missing interaction among membership variables. In this proposed algorithm, highly and lightly expressed data points are separated more efficiently due to covariance pressure. The algorithm is described and compared with the most elevated techniques such as k-means (KM), fuzzy c-means (FCM), and penalized fuzzy c-means (PFCM) clustering techniques. These techniques are verified for different validity measures for artificial dataset and a Brain Tumor gene expression dataset. Our proposed clustering algorithm shows a much higher usability than the other related techniques.