The Influence of Induced OWA Operators in a Clustering Method

Abstract

In this work we present an adaptation of the well known k-means algorithm for clustering. The proposal increases the flexibility of the algorithm to calculate the representative value of each cluster. To do so, we work with Induced Ordered Weighting Averaging operators. These instances of aggregation functions are able to increase or decrease the influence of the data in the final result depending on the specific values of the weights. We present an experimental study to show how these operators are able to modify the representatives of the clusters. We also compare our results over some standard datasets.