A genetic algorithm-based framework for wavelength selection on sample categorization.

Area of Science:

• Chemometrics and Spectroscopy
• Forensic Science and Pharmaceutical Analysis

Background:

• Forensic and pharmaceutical analyses rely on chemometrics and optimization for identifying seized drug and medicine samples.
• Accurate classification of samples (e.g., authentic vs. forged medicines, cocaine salt vs. base form) is crucial for tracking illegal operations.

Purpose of the Study:

• To propose a novel framework for identifying relevant subsets of attenuated total reflectance Fourier transform infrared (ATR-FTIR) wavelengths.
• To classify samples into distinct categories, improving accuracy and efficiency in forensic and pharmaceutical investigations.

Main Methods:

• ATR-FTIR spectra were partitioned into equidistant intervals, with k-nearest neighbour (KNN) classification applied to each interval.
• Selected intervals were refined using a genetic algorithm (GA) to identify a limited number of wavelengths that maximize classification accuracy.
• The framework was tested on Cialis®, Viagra®, and cocaine ATR-FTIR datasets.

Main Results:

• The proposed method significantly reduced the number of required wavelengths for sample categorization.
• Classification accuracy was substantially increased across the tested datasets.
• The GA-based approach provided more refined wavelength selection compared to traditional interval methods.

Conclusions:

• The novel framework offers valuable information for monitoring illegal drug and medicine production.
• Reduced wavelength subsets enable the development of portable devices for rapid, on-site authenticity testing.
• This approach minimizes reliance on expensive and time-consuming laboratory analyses.