Data-driven modeling of water adsorption isotherms in cocoa beans: Dataset and Python-based Machine Learning tools for multivariate analysis and storage management

This work presents a comprehensive dataset of water adsorption isotherms for dried and roasted cocoa beans ( Theobroma cacao L.), complemented by a set of Python-based machine learning (ML) tools designed to support data-driven multivariate modeling and storage management. Adsorption isotherms were experimentally determined using the Dynamic Dewpoint Isotherm (DDI) method in the range of water activities (a w ) of 0.1-0.85 and at temperatures representative of post-harvest storage facilities (25, 30, and 40°C). The water adsorption isotherms provided a detailed analysis of the hygroscopic behavior of dried and roasted cocoa beans, enabling the assessment of moisture-related stability during storage. The dataset includes Excel files containing equilibrium moisture content (X e ) values, experimental conditions (a w , temperature and type of cocoa beans) and replicate measurements, thereby enabling reproducible and traceable modeling workflows. To complement the experimental data, this work also provides fully documented Python scripts for multivariate mathematical modeling and prediction of water adsorption behavior using ML techniques. The computational ML tools were developed in Python using the Spyder integrated development environment (IDE) provided by the Anaconda distribution, with libraries such as scikit-learn and pandas . These tools implement automated routines for model calibration and grid search-based hyperparameter optimization of Support Vector Machine (SVM), Random Forest (RF), and Artificial Neural Networks (ANN). These models enable the prediction of X e as a function of a w , temperature, and cocoa processing type (dried or roasted), supporting robust multivariate modeling strategies. The integrated dataset and Python workflow together constitute a methodological framework for predicting moisture behavior in cocoa during storage, assessing storage-related risks based on model outputs, and supporting informed decision-making throughout the cocoa supply chain. By enabling reliable estimation of moisture-related storage parameters in cocoa, critical aspects associated with stability, spoilage prevention, and shelf-life determination can be addressed. Consequently, this work provides a valuable reference dataset for researchers, cocoa producers, and industry stakeholders seeking to optimize storage conditions and maintain cocoa quality during post-harvest handling.