Use of a complex classification system in pattern recognition problems for the determination of structure-activity relations

Abstract

A complex classification system consisting of autoscaling (fitting the parameters used to a mean value of 0 and a variance of 1), a 2nd linear transformation of characters which yields the uncorrelated properties, and a cluster anal. was used in the pattern recognition procedure for detg. structure-activity relations, as illustrated by a series of 106 di-Ph ethers and their herbicidal activity.