No abstract

No abstract

No abstract

Cocaine sample correlation provides important information in the identification of traffic networks. However, available methods for estimating if samples are linked or not require the use of previous police investigation and forensic expert knowledge regarding the number of classes and provide thresholds that are both static and data set specific. In this paper, a novel unsupervised linkage threshold method (ULT) based on chemometric analysis is described and applied to the analysis of headspace gas chromatography mass spectrometry (HS-GC/MS) data of more than 250 real cocaine hydrochloride samples seized by Brazilian Federal Police. The method is capable of establishing linkage thresholds that do not require any prior information about the number of classes or distribution of the samples and can be dynamically updated as the data set changes. It is envisaged that the ULT method may also be applied to other forensic expertise areas where limited population knowledge is available and data sets are continually modified with the inflow of new information.

No abstract