Statistical tree classification of aphids based on morphological characteristics

Abstract

Individual aphids were classified to their clones and separated to their host groups using a novel non-parametric classification tree method. The classification of the individuals was based on morphometric variables measured for each individual. The classification tree method splits the initial set of individuals recursively into subsets using one of the variables. The method has the form of a tree branching off into intermediate terminal nodes. The splitting criterion is the increase in purity when the node is split into two subnodes. The size of the tree is controlled by a threshold level for the improvement of the apparent misclassification rate of the tree after each splitting step. The results obtained by applying the classification tree method were in good agreement with those obtained by the conventional discriminant methods such as Fisher's linear discriminant analysis and canonical variate analysis. The classification tree method has the advantage over the other two discriminant methods that it gives a graphical presentation of the structure of the data at any growing stage of the tree. Therefore, it can classify the aphids into their clones and, at the same time, to separate the host groups using a single tree. (C) 1999 Elsevier Science B.V. All rights reserved.