Decisions Fusion Strategy: Towards Hybrid Cluster Ensemble

Abstract

Clustering ensembles have renowned as a powerful method for improving both the performance and constancy of unsupervised classification solutions. However, finding a consensus clustering from multiple algorithms is a difficult problem that can be approached from combinatorial or statistical perspectives. We offer a new clustering strategy which is formulated to cluster extracted mammography features into soft clusters using unsupervised learning strategies and 'fuse' the decisions using majority voting and parallel fusion in conjunction with a neural classifier. The idea is to observe associations in the features and fuse the decisions (made by learning algorithms) to find the strong clusters which can make impact on overall classification accuracy. Two novel techniques are proposed for fusion, majority-voting based data fusion, and neural-based fusion. The proposed approaches are tested and evaluated on the benchmark database-digital database for screening mammograms (DDSM). This study compares the performance of the proposed ensemble approach with other fusion approaches for clustering ensembles. Experimental results demonstrate the effectiveness of the proposed method on benchmark dataset.

Selective Clustering Ensemble Based on Covariance

“…Luo et al [6] used information theory to design a genetic algorithm to combine multiple clusterings. Hassan et al [7] developed a ensemble method with majority voting and parallel fusion in conjunction with a neural classifier. Mohammadi et al [8] stated an evolutionary approach to clustering ensemble, and they used an evolutionary combinational clustering method to find the number of clusters.…”