Optimization of Formulation and Chemical Quality as well as Antioxidant Activity of Honey Powder Using Adaptive Feature Selection Multi-Target Learning (AFS MTL) Cindy Raficha Sari (a), Firman Jaya (b), Rizal Setya Perdana (c), Ekasari Nugraheni (d), Dianadewi Riswantini (d), Rika Sustika (d), Agus Susilo (b), Osfar Sjofjan (e)
a) Faculty of Animal Science, Universitas Brawijaya, Malang 65145, Indonesia
*author[at]ub.ac.id
b) Department of Animal Products Technology, Faculty of Animal Science,
Universitas Brawijaya, Malang 65145, Indonesia
c) Information System Department, Computer Science Faculty, Universitas
Brawijaya, Malang 65145, Indonesia
d.) Research Center for Information and Data Sciences, National Research and
Innovation Agency, Bandung, Indonesia
e.) Department of Feed and Animal Nutrition, Faculty of Animal Science,
Universitas Brawijaya, Malang 65145, Indonesia
Abstract
This study aims to develop a predictive system based on Adaptive Feature
Selection Multi-Target Learning (AFS-MTL) to optimize powdered honey
formulation through the selection of the best features from variables such as
honey content, type and content of fillers, drying method, temperature, time, and supporting additives. Three machine learning algorithms, namely Random Forest, Support Vector Machine (SVM), and XGBoost, were used to predict the optimal features and powdered honey formulation for chemical quality parameters (moisture content, water activity, HMF, reducing sugar, diastase enzyme, and sugar composition) and antioxidant activity (DPPH, ABTS, total phenols, and total flavonoids). The main ingredients used in this study included Acacia monoflora honey that had undergone evaporation and pasteurization to achieve a moisture content of 20%, fillers such as maltodextrin and gum arabic, and anti-caking agents. The predictive performance of AFS-MTL was evaluated using the Average Correlation Coefficient (aCC) metric to measure the suitability of the prediction results with the actual data. This algorithm has great potential for accurate prediction of powdered honey quality parameters and efficient optimal formulation recommendations.
