Synergizing genetic algorithms and neural networks to decode mineral-antioxidant dynamics in black-raspberry

The applications of artificial intelligence (AI) in complex agricultural systems offers innovative opportunities to improve the yields of bioactive compounds through effective nutrient management strategies. This study leverages AI and machine learning to investigate relationships between mineral composition and antioxidant activity in black raspberry ( Rubus occidentalis ), a critical crop for functional food markets. By combining biochemical profiling with advanced computational models, we map interactions among minerals (P, K, Ca, Mg, Cu, Fe, Mn) and antioxidant metrics (DPPH, FRAP). Feature selection via principal component analysis (PCA), mutual information (MI) and correlation analysis prioritized P, K, Ca, Mg, and Cu as dominant factors, with Mg and Cu showing strong positive associations (p < 0.01) with antioxidant activity. The multi-layer perceptron (MLP) model demonstrated superior accuracy in predicting DPPH (testing R² = 0.72, RMSE = 0.98) and FRAP (testing R² = 0.92, RMSE = 0.54), outperforming general regression neural networks (GRNN), support vector regression (SVR), and multiple linear regression (MLR). Hybrid MLP-Genetic Algorithm optimization identified that reducing potassium inputs by 30 % while elevating magnesium and copper by 22 % and 13 %, respectively, could amplify radical scavenging capacity (54 %) and reducing power (89 %). This study establishes an innovative AI-driven methodology for optimizing mineral management in agriculture, directly linking computational insights to actionable farming strategies to enhance bioactive compound production, thereby advancing sustainable crop solutions for functional food markets.