Discriminative feature analysis of dairy products based on machine learning algorithms and Raman spectroscopy

Area of Science:

• Food Science
• Analytical Chemistry
• Spectroscopy

Background:

• Food quality control relies on accurate discriminant analysis of similar samples.
• Raman spectroscopy and machine learning offer a powerful approach for intelligent food discrimination.
• Understanding spectral feature patterns is crucial for enhancing discrimination techniques.

Purpose of the Study:

• To investigate Raman spectral features for discriminating three brands of dairy products.
• To evaluate the performance of Support Vector Machines (SVM), Extreme Learning Machines (ELM), and Convolutional Neural Networks (CNN) algorithms.
• To determine the impact of selected spectral feature intervals on recognition accuracy and computational efficiency.

Main Methods:

• Utilized Raman spectroscopy to collect spectral data from dairy product samples.
• Applied machine learning algorithms including SVM, ELM, and CNN for data analysis.
• Performed feature spectral analysis to identify optimal spectral intervals for discrimination.
• Analyzed sample distribution using Euclidean distance based on feature spectra.

Main Results:

• Optimal spectral feature intervals varied among SVM, ELM, and CNN algorithms.
• High recognition accuracy (100%) was achieved by all tested algorithms with specific spectral ranges.
• ELM demonstrated superior computational efficiency (<0.3s) compared to SVM (200s) and CNN (80s).
• Visual analysis of spectral feature intervals provided insights into sample distribution.

Conclusions:

• Selecting appropriate spectral feature intervals is key to balancing recognition accuracy and computational efficiency.
• Different machine learning algorithms benefit from distinct spectral feature ranges for optimal performance.
• This study provides a strategic framework for analyzing spectral features in discriminant research for food quality control.