Data-Driven Approaches to Coffee Quality Prediction: A Research Synthesis Danang Triagus Setiyawan 12*, Moses Laksono Singgih 2, I Ketut Gunarta 2
1 Department of Agroindustrial Technology, Faculty of Agricultural Technology, Universitas Brawijaya, Jl. Veteran, Malang 65145, Indonesia
2 Department of Industrial and Systems Engineering, Institut Teknologi Sepuluh Nopember (ITS), Jl. Raya ITS, Surabaya 60111, Indonesia
*Email: danangtriagus[at]ub.ac.id
Abstract
This study provides a comprehensive synthesis of research on data-driven approaches for predicting coffee quality. The review highlights how diverse machine learning methods have been applied to generate reliable predictions, improve decision-making, and support quality management practices. The synthesis covers issues such as handling imbalanced datasets, probability calibration, evaluation metrics, and the balance between interpretability and predictive performance. Comparative findings across methods including logistic regression, random forest, and gradient boosting are discussed in terms of discrimination power, calibration strength, and practical deployment. The results demonstrate that data-driven models can offer both accuracy and transparency when appropriately designed, making them valuable tools for routine coffee quality assessment and management.
