Clustering honey samples with unsupervised machine learning methods using FTIR data

Area of Science:

• Food Science
• Analytical Chemistry
• Spectroscopy

Background:

• Honey authentication and classification are crucial for quality control.
• Spectral analysis offers a non-destructive method for characterizing food products.
• Fourier Transform Infrared (FTIR) spectroscopy provides rich chemical information.

Purpose of the Study:

• To develop an efficient method for clustering and categorizing honey samples using FTIR spectral data.
• To assess the accuracy of a deep learning model for honey classification.
• To demonstrate a practical approach for honey analysis with minimal preprocessing.

Main Methods:

• Fourier Transform Infrared (FTIR) spectroscopy for data acquisition.
• Elbow method for determining the optimal number of clusters (five identified).
• Principal Component Analysis (PCA) for dimensionality reduction.
• Hierarchical Cluster Analysis (HCA) for cluster refinement.
• Deep learning with a Multilayer Perceptron (MLP) for classification.

Main Results:

• Successful clustering of honey samples into five distinct groups.
• High classification accuracy of 96.15% achieved with the MLP model.
• Demonstrated reliability and efficiency of the FTIR-based approach.

Conclusions:

• FTIR spectral data combined with machine learning provides a powerful tool for honey sample categorization.
• The proposed method is swift, cost-effective, and requires minimal sample preprocessing.
• This approach shows significant promise for spectral analysis applications in food science.