A new clustering scheme for crisp data based on a membership function and owa operator

Clustering is a very important tool which is applied in several areas, ranging from pattern recognition and marketing to chemistry. A majority of the clustering algorithms classify observations based on distance measures. According to the literature, if the units of measurement of the variables are different, then the result of the clustering is said to be unreliable. Even sometimes, distance based clustering shows contradictory results when measurement units are closely related. Therefore, a new clustering scheme is proposed in this paper based on combining the membership function and OWA operator when classic clustering seems to have failed. For this purpose, a real data set from chemistry with ten variables are used to exemplify the new clustering scheme.