A machine learning-assisted colorimetric sensor array based on Fe-doped MnO₂/graphene composites for flavonoid identification and bioactivity prediction in vine tea.

Accurate characterization of flavonoids in vine tea is of considerable importance for assessing its nutritional value and exploring its potential applications in hypoglycemic and hypolipidemic agents. Combining the substrate-responsive catalytic activity of MnFe/GO nanozymes with machine learning-based pattern recognition enabled the rapid and accurate differentiation of three structurally similar flavonoids. Furthermore, a two-stage cascade prediction model was constructed: the first stage employed random forest and gradient boosting regression to predict total concentration and component molar ratios; the second stage used the predicted composition vector as input to a random forest multi-output regression model to predict α-glucosidase and pancreatic lipase inhibitory rates. The results demonstrated good predictive performance and generalization ability. The model demonstrated satisfactory performance in real vine tea samples, enabling simultaneous prediction of flavonoid composition and bioactivity, thereby providing an effective approach for rapid quality assessment of natural products.