Bio-inspired deep learning framework with intelligent optimization for enhanced yield prediction in smart aeroponic cultivation

Aeroponic farming enables high-efficiency plant production through precise environmental control; however, predicting crop yield remains challenging due to nonlinear interactions among sensor variables and operational parameters. The objective of this study is to develop an intelligent and robust yield-prediction framework capable of improving decision support in controlled aeroponic systems. Existing approaches often rely on limited feature selection and manually tuned learning models, which restrict prediction accuracy, scalability, and interpretability. To address this gap, we propose a hybrid framework that integrates Fig Wasp Symbiotic Co-evolution (FWSC)–based feature selection, a Capsule Network (CapsNet) for hierarchical representation learning, and Fennec Fox Optimization (FFO) for adaptive hyperparameter tuning. The framework was evaluated using a real-world aeroponic dataset containing environmental, reservoir, and operational sensor measurements from multiple crop cycles. Descriptive statistics and distribution analyses were applied during preprocessing, and model performance was assessed using standard regression and classification metrics. The proposed model achieved superior predictive performance compared with baseline approaches, obtaining an R 2 of 0.983, 99.13% accuracy, and a significantly reduced prediction error. SHAP-based interpretability analysis revealed that TDS, reservoir temperature, and CO 2 concentration were among the most influential features affecting yield. The results demonstrate that the FWSC–CapsNet–FFO framework enhances learning efficiency, generalization capability, and model transparency for aeroponic yield prediction. Future work will focus on extending the framework to multi-site cultivation environments, integrating multimodal sensing (including imaging data), and deploying lightweight model variants for real-time, IoT-enabled precision agriculture applications.