Tutorial paper XV: Classification automatique

Abstract

Very diverse procedures have been proposed in order to separate a set of objects into classes. This paper presents an overview of the methods. Firstly, one considers the hierarchical methods for metric or non-metric data with agglomerative or divisive algorithms. Then one recalls some non-hierarchical methods the resulting clusters of which form a partition or an overlapping partition or a fuzzy partition. Finally, one gives various criteria for the assessment of classifications produced by all these methods.