Evidence Accumulation Clustering Based on the K-Means Algorithm

Abstract

Abstract. The idea of evidence accumulation for the combination of multiple clusterings was recently proposed [7]. Taking the K-means as the basic algorithm for the decomposition of data into a large number, k, of compact clusters, evidence on pattern association is accumulated, by a voting mechanism, over multiple clusterings obtained by random initializations of the K-means algorithm. This produces a mapping of the clusterings into a new similarity measure between patterns. The final data partition is obtained by applying the single-link method over this similarity matrix. In this paper we further explore and extend this idea, by proposing: (a) the combination of multiple K-means clusterings using variable k; (b) using cluster lifetime as the criterion for extracting the final clusters; and (c) the adaptation of this approach to string patterns. This leads to a more robust clustering technique, with fewer design parameters than the previous approach and potential applications in a wider range of problems.

Evidence Accumulation Clustering with Possibilitic Fuzzy C-Means base clustering approach to disease diagnosis

“…The Evidence Accumulation Clustering (EAC) [21][22][23] is an ensemble clustering method for combining multiple clustering approaches in order to achieve better performance than is obtainable using single clustering methods. Ensemble clustering uses a consensus of several clustering solutions and merges them into a single consensus solution, so that improved robustness and stability can be achieved in comparison with the single clustering method [18,19].…”

“…Ensemble clustering uses a consensus of several clustering solutions and merges them into a single consensus solution, so that improved robustness and stability can be achieved in comparison with the single clustering method [18,19]. There are several approaches to accumulate evidence of clustering methods: the first involves combining the results of different clustering algorithms; the second requires resampling the data using different techniques, such as bagging and boosting, so that different results are created; and the third applies a clustering algorithm many times, each with different initialisation [21][22][23]. The EAC ensemble method, as proposed by Fred and Jain in [21][22][23], uses the third approach by applying HCM algorithms, such as K-Means algorithm, which will henceforth be referred to simply as the hEAC technique, as the underlying clustering algorithm to produce clustering ensembles, as explained in the sequel.…”

“…The hEAC ensemble clustering method is executed in three broad steps or processes, namely: the splitting, combination, and merging processes [21][22][23]. In the first step, the data is split into a large number of clusters and different partitions are obtained by random initialisations of the Kmeans algorithm.…”

“…However, finding a consensus cluster from multiple partitions is still a difficult problem that has been approached from different angles including those that are graph-based, combinatorial or statistical in nature [19][20][21]. The Evidence Accumulation Clustering (EAC) method, proposed by Fred and Jain [22][23], employs a Hard C-Means (HCM) method (notably, the K-Means algorithm) to map the individual data partitions in a clustering ensemble into a new similarity measure between patterns [23]. A typical EAC ensemble method based on HCM base clustering (or simply hEAC ensemble technique) is executed in three stages: namely, the splitting, combination, and merging processes or steps [22][23].…”

“…The Evidence Accumulation Clustering (EAC) method, proposed by Fred and Jain [22][23], employs a Hard C-Means (HCM) method (notably, the K-Means algorithm) to map the individual data partitions in a clustering ensemble into a new similarity measure between patterns [23]. A typical EAC ensemble method based on HCM base clustering (or simply hEAC ensemble technique) is executed in three stages: namely, the splitting, combination, and merging processes or steps [22][23]. However, with its tight and well-defined boundaries, the hEAC clustering technique suffers from lack of flexibility in terms of membership of clusters and the impact resulting therefrom on achieving the convergence needed to produce a stable combined clustering outcomes [24].…”