Comparative assessment of the modeling and discrimination power of two pattern recognition methods applied to detect designer drugs

Abstract

We are presenting a comparative assessment of the modeling and discrimination power of two pattern recognition methods, i.e. Hierarchical Cluster Analysis (HCA) and the Naive Bayes Classifier (NBC), from the point of view of their efficiency in detecting illicit amphetamines, based on their GC-IRAS laser spectra recorded between 1405 and 1150 cm-1. A special attention was also given to the detection of their main precursors, the ephedrines. The spectra were first preprocessed with a discriminating feature weight wTE. The performances of two automatic detection applications, based on HCA and on NBC, are compared from the point of view of their capacity to correctly recognize illicit amphetamines and ephedrines and distinguish among them according to the Schedules of the United Nations Convention on Psychotropic Substances.