Cluster Ensemble Based on Co-occurrence Data

Abstract

Ensemble approach based on aggregated models has been successfully applied in the context of supervised learning in order to increase the accuracy and stability of classification. Recently, analogous techniques for cluster analysis have been suggested. Research has proved that, by combining a set of different clusterings, an improved solution can be obtained. In the traditional way of learning from a data set, the classifiers are built in a feature space. However, an alternative way can be found by constructing decision rules on dissimilarity representations. In such a recognition process each object is described by a matrix showing the similarities or distances to the rest of training samples. This research has focused on exploiting the additional information provided by a collection of diverse clusterings to generate a co-occurrence (co-association) matrix. Taking the co-occurrences of pairs of patterns in the same cluster as votes for their association, the data partitions are mapped into a co-association matrix. This n ×n matrix represents a new similarity measure between patterns. The final data partition is obtained by clustering this matrix. In the experiments, the behavior of partitions built on co-occurrence data is studied.