Application of Revised Version of Neural Independent Component Analysis to Classification Problems of Confiscated Methamphetamine

Abstract

Recently, profiling the chemical substances in illegally distributed drugs has been needed in order to reveal the drug channels. However, this kind of profiling is often difficult because such drugs contain various kinds of impurities and the quantity of these impurities changes. Due to these circumstances, several methods, including a slightly revised ICA (Independent Component Analysis) by a Hebbian learning artificial neural network, were applied for profiling illegally distributed methamphetamine. Eventually, better classification results with the ICA than with other methods were obtained. These results show that ICA could make it easier to profile illegally distributed methamphetamine.