Cluster validation techniques for genome expression data

Several clustering algorithms have been suggested to analyse genome expression data, but fewer solutions have been implemented to guide the design of clustering-based experiments and assess the quality of their outcomes. A cluster validity framework provides insights into the problem of predicting the correct the number of clusters. This paper presents several validation techniques for gene expression data analysis. Normalisation and validity aggregation strategies are proposed to improve the prediction about the number of relevant clusters. The results obtained indicate that this systematic evaluation approach may significantly support genome expression analyses for knowledge discovery applications. (C) 2002 Elsevier Science B.V. All rights reserved.