Standardization of UV-visible data in a food adulteration classification problem

Area of Science:

• Analytical Chemistry
• Chemometrics
• Food Safety

Background:

• Spectral variations due to experimental conditions can degrade multivariate classification models.
• Adulteration of culinary spices with banned dyes like Sudan I-IV poses a significant food safety concern.
• UV-visible spectroscopy combined with chemometric methods offers a viable approach for spice analysis.

Purpose of the Study:

• To evaluate multivariate calibration transfer methods for classification tasks.
• To assess the effectiveness of Piecewise Direct Standardization (PDS) in mitigating spectral variation.
• To accurately classify culinary spices adulterated with Sudan dyes.

Main Methods:

• Utilized UV-visible spectroscopy for sample characterization.
• Applied Partial Least Squares-Discriminant Analysis (PLS-DA) for classification.
• Implemented Piecewise Direct Standardization (PDS) for spectral data standardization.

Main Results:

• Standardization using PDS successfully addressed spectral variations.
• The PDS-processed data yielded classification results comparable to the initial high-performing PLS-DA model.
• Accurate discrimination between unadulterated and adulterated spice samples was achieved.

Conclusions:

• Piecewise Direct Standardization is an effective method for improving multivariate classification performance in the presence of spectral variation.
• The study demonstrates a robust approach for detecting banned dye adulteration in culinary spices.
• Standardization techniques are crucial for maintaining reliable analytical model performance across different experimental conditions.