Computer-aided classification techniques for biological taxonomy

A computer-aided classification algorithm is basically a procedure for clustering a large number of objects into subsets representing similar "types." The cluster analysis technique is used to select objects which have similar characteristics and is distinct from such techniques as discriminant analysis and other more recently developed pattern recognition methods in the sense that it is not assumed a priori which objects belong to specific groups or what probability distributions characterize the different groups. The primary value of cluster analysis is in the classification of objects into "natural" groups as suggested by the data itself. This paper discusses recently developed approaches to three fundamental problems associated with the creation of computer-aided classification techniques for taxonomic studies of biological populations: 1) Data Characterization; 2) Clustering Procedure, and 3) Statistical Significance of Clustering Results, A brief presentation of biological applications is included.