Abstract
Efficient irrigation management in agriculture is key to optimizing water use, ensuring optimal crop health and productivity. Existing machine learning (ML) models demonstrate insufficient classification accuracy in classifying irrigation needs for various crops, motivating new methods. This study identifies a gap in using bio-inspired optimization algorithms and ML to enhance irrigation classification. This research develops a hybrid intelligent model combining the improved grey wolf optimization (IGWO) and artificial neural networks (ANN) to classify irrigation needs for different crops. IGWO employs an adaptive weighted position update strategy that assigns greater influence to the best solution during ANN weight initialization, effectively avoiding local minima traps and providing more consistent near-optimal parameter configurations compared to standard optimization approaches. The IGWO-ANN model is trained using crucial soil and environmental variables including soil moisture, air temperature, humidity, and soil temperature. The proposed model achieves a mean accuracy of 98.72% ± 0.44%, with precision, recall, F1-score, and AUC of 93.46% ± 1.32%, 99.43% ± 1.47%, 96.35% ± 1.25%, and 0.9994 ± 0.0009 respectively, evaluated across 10 independent runs to ensure statistical reliability. Additionally, the gradient-based fine-tuning phase requires only 0.44 s, indicating computational efficiency. The proposed model is benchmarked against 10 competing models including metaheuristic variants, deep learning architectures, and ensemble methods, demonstrating superior and statistically stable classification performance. The results reveal that the proposed IGWO-ANN approach enhances prediction reliability and reduces the limitations of standalone ANN models. Thus, the proposed model can significantly reduce water loss during the irrigation process and contribute toward sustainable agriculture and efficient water resource management.