Dataset and machine learning-based computer-aided tools for modeling working sorption isotherms in dried parchment...

Area of Science:

• Agricultural Science
• Food Science
• Analytical Chemistry

Background:

• Understanding water sorption is crucial for coffee bean storage and quality.
• Parchment husk influences the hygroscopic behavior of coffee beans.
• Mid-infrared spectroscopy offers insights into coffee composition and properties.

Purpose of the Study:

• To present a comprehensive dataset of working sorption isotherms and mid-infrared spectra for parchment husk, parchment coffee, and green coffee beans.
• To develop computer-aided tools for mathematical modeling of sorption isotherms and infrared data using machine learning.
• To provide practical tools for optimizing coffee storage and quality monitoring.

Main Methods:

• Working sorption isotherms determined using the Dynamic Dewpoint Isotherm (DDI) method across various water activity (a_w) and temperature conditions.
• Mid-infrared spectra acquired using Attenuated Total Reflectance-Fourier Transform Infrared (ATR-FTIR) spectroscopy.
• MATLAB scripts developed for Support Vector Machine (SVM), Random Forest (RF), and Principal Component Analysis (PCA) modeling.

Main Results:

• Experimental data on water sorption isotherms and ATR-FTIR spectra for different coffee types under controlled conditions.
• Validated machine learning models (SVM, RF) for predicting equilibrium moisture content (X_e) based on a_w, temperature, and coffee type.
• PCA models for robust differentiation of coffee samples using infrared spectral data.

Conclusions:

• The dataset and computational tools facilitate a deeper understanding of coffee's hygroscopic behavior.
• Optimized storage conditions can be determined, leading to improved shelf-life and quality.
• The study provides valuable resources for researchers, producers, and stakeholders in the coffee industry.