Application of bio-inspired methods within cluster forest algorithm

Abstract

Cluster Forest (CF) is relatively new ensemble clustering method inspired by Random Forest algorithm. The main idea behind of the existing algorithm consists in a construction of a larger number of partial clusterings for feature subsets using K-means algorithm. At the end, these clusterings are aggregated using a method of spectral clustering. This article describes a new application of bio-inspired methods that replaces the K-means algorithm in the computation pipeline. Several bio-inspired methods were tested on eight different datasets and compared with the original CF and others well known clustering methods.