New Zealand Statistical Association 2024 Conference

Yongshi Deng

ESR

Poster display: Cluster analysis of unknown samples using Lumi™ drug scan data

This is joint work with Dion Sheppard

Lumi™ Drug Scan provides real-time detection of illicit drugs through a portable Near-infrared device. The initial deployed model identified street samples as methamphetamine, MDMA, cocaine, or unknown. A large portion of samples scanned by NZ frontline police officers falls into the unknown category. It is of interest to uncover what other popular substances may constitute these unknown samples. This analysis can provide meaningful descriptive statistics of the street samples seized by the police and offer valuable insights for future model development. In this poster, we will demonstrate how semi-supervised clustering can be used to achieve these goals.