Open Access Te Herenga Waka-Victoria University of Wellington
Browse

Genetic Programming for Symbolic Regression with Portable Near-Infrared Spectroscopy for the Prevention of Harm from Adulterated Illicit Drugs at Festivals in New Zealand

Download (2.73 MB)
thesis
posted on 2025-05-06, 21:15 authored by Steven Dockter

Illicit drugs are often mixed with cutting agents and adulterants, which can pose significant risks, either independently or when consumed with the drug itself. Drug checking services play a vital role in reducing these risks by providing information about substance composition and drug harm prevention. A widely adopted model for delivering drug check- ing services is at events and festivals. To effectively reduce drug-related harm in such settings, it is crucial to have rapid, accurate, non-destructive, safe, portable, and user-friendly substance identification methods. Near- infrared spectroscopy (NIRS) is a well-established technique for identi- fying and quantifying a wide range of substances, including illicit drug mixtures, particularly with high-end bench-top instruments. While using portable NIRS devices for mixture analysis is advantageous, challenges such as low sensitivity, limited wavelength range, and low resolution per- sist. To address these challenges, various artificial intelligence techniques have been applied to analyze portable NIRS data for mixtures. However, evolutionary computation methods, which are known for their robustness in solving complex optimization problems, have not been widely explored in NIRS-based mixture analysis. This thesis proposes a genetic program- ming for symbolic regression (GPSR) approach to develop explainable models for analyzing illicit drug mixtures using portable NIRS. Exper- imental results demonstrate that the proposed approach can accurately model the relationship between a mixture and its components based on NIRS data. Furthermore, the method shows promise in accurately iden- tifying components within a drug sample, with performance comparable to traditional linear models. The potential for implementing this method into an integrated solution for drug checking services at point-of-care sce- narios, such as festivals, is also explored.

History

Copyright Date

2025-05-07

Date of Award

2025-05-07

Publisher

Te Herenga Waka—Victoria University of Wellington

Rights License

CC BY 4.0

Degree Discipline

Artificial Intelligence

Degree Grantor

Te Herenga Waka—Victoria University of Wellington

Degree Level

Masters

Degree Name

Master of Science

ANZSRC Socio-Economic Outcome code

280115 Expanding knowledge in the information and computing sciences

ANZSRC Type Of Activity code

3 Applied research

Victoria University of Wellington Item Type

Awarded Research Masters Thesis

Language

en_NZ

Alternative Language

en

Victoria University of Wellington School

School of Engineering and Computer Science

Advisors

Chen, Qi