Prediction of Powdered Honey Formulation Based on Functional Properties with Ensemble Learning in Production Scenario Modeling (EL-PSM)

Powdered honey offers advantages, especially for the industrial sector, because it is more practical in terms of storage and distribution. The formulation of powdered honey is highly dependent on the type of carrier, mixing ratio, and drying conditions. This study aims to determine the optimal powdered honey formulation through the advantages of applying the XGBoost algorithm, which has tolerance for missing data, Random Forest, which is capable of working with large data sets, and Support Vector Machine, which is capable of working even with messy data. The Ensemble Learning technique will then follow this in Production Scenario Modeling (EL-PSM). This method is used to improve prediction accuracy and model reliability in identifying the best combination of honey and carrier materials. Experimental data generated from various formulation variations are processed using the three algorithms, then combined through an ensemble approach to obtain a more robust model. The results show that the application of EL-PSM more accurate, stable, and consistent predictions of powdered honey formulations compared to the use of a single algorithm. The research demonstrates that integrating machine learning with a production scenario approach can be an effective strategy for developing honey-based food products with enhanced functional properties and efficiency.