Toxic activity evaluation by clustering

Abstract

A toxic activity data set has been processed by crisp partitioning clustering. A validation procedure to evaluate the number of clusters is presented. The procedures is based on the use of SOM and k–means algorithms. The resulting clustering method is based on the Mutual Information index, whose task is to identify the optimal number of clusters in order to obtain most reliable clustering results. Specifically, this method has been applied to the study of the phytotoxic activity of Salvia genus plants, taking into account only the most significant species to reduce costs. In this respect the use of raking methods in order to provide a reliable classification is important.