I. Batyrshin and G. Sidorov (Eds.): MICAI 2011, Part II, LNAI 7095, pp. 153–164, 2011.
© Springer-Verlag Berlin Heidelberg 2011
Clustering Ensemble Framework via Ant Colony
Hamid Parvin and Akram Beigi
Islamic Azad University, Nourabad Mamasani Branch, Nourabad Mamasani, Iran
hamidparvin@mamasaniiau.ac.ir, beigi@iust.ac.ir
Abstract. Ensemble-based learning is a very promising option to reach a robust
partition. Due to covering the faults of each other, the classifiers existing in the
ensemble can do the classification task jointly more reliable than each of them.
Generating a set of primary partitions that are different from each other, and
then aggregation the partitions via a consensus function to generate the final
partition, is the common policy of ensembles. Another alternative in the ensem-
ble learning is to turn to fusion of different data from originally different
sources. Swarm intelligence is also a new topic where the simple agents work in
such a way that a complex behavior can be emerged. Ant colony algorithm is a
powerful example of swarm intelligence. In this paper we introduce a new en-
semble learning based on the ant colony clustering algorithm. Experimental re-
sults on some real-world datasets are presented to demonstrate the effectiveness
of the proposed method in generating the final partition.
Keywords: Ant Colony, Data Fusion, Clustering.
1 Introduction
Data clustering is an important technique for statistical data analysis. Machine learn-
ing typically regards data clustering as a form of unsupervised learning. The aim of
clustering is the classification of similar objects into different cluster, or partitioning
of a set of unlabeled objects into homogeneous groups or clusters (Faceli et al., 2006).
There are many applications which use clustering techniques to discover structures in
data, such as Data Mining (Faceli et al., 2006), pattern recognition, image analysis,
and machine learning (Deneubourg et al., 1991).
Ant clustering is introduced by Deneubourg et al. (1991). In that model, the swarm
intelligence of real ants is inserted into a robot for the object collecting task. Lumer
and Faieta (1994) based on how ants organize their food in their nest, added the Euc-
lidean distance formula as similarity density function to Deneubourg’s model. Ants in
their model had three kinds of abilities: speed, short-term memory, and behavior ex-
change.
There are two major operations in ant clustering: picking up an object from a clus-
ter and dropping it off into another cluster (Tsang and Kwong, 2006). At each step,
some ants perform pick-up and drop-off based on some notions of similarity between
an object and the clusters. Azimi et al. (2009) define a similarity measure based on the
co-association matrix. Their approach is fully decentralized and self-organized and
allows clustering structure to emerge automatically from the data.