|Clustering is an established unsupervised learning method. Substantial research has been carried out in the area of feature weighting, as well instance selection for clustering. Some work has paid attention to instance weighted clustering algorithms using various instance weighting metrics based on distance information, geometric information and entropy information. However, little research has made use of instance density information to weight instances. In this paper we use density to define instance weights. We propose two novel instance weighted clustering algorithms based on Local Outlier Factor and compare them against plain k-means and traditional instance selection.|
*** Title, author list and abstract as seen in the Camera-Ready version of the paper that was provided to Conference Committee. Small changes that may have occurred during processing by Springer may not appear in this window.