A comparative analysis of optimized CNN models using bio-inspired algorithms

Nitrogen is a macronutrient that is responsible for crop development. Accurate nitrogen prediction improves crop productivity. This study aims to enhance nitrogen prediction in wheat crops using convolutional neural networks (CNN) with bio-inspired algorithms. CNN, which is widely used for image classification, relies on carefully chosen parameters such as the number of layers, learning rate and kernel sizes to perform effectively. These parameters were optimized using five bio-inspired algorithms. Optimization algorithms aid in selecting the parameters of CNNs to make them more accurate and efficient. The algorithms are genetic algorithms grey wolf optimizer, particle swarm optimization, ant bee colony, and evolutionary algorithms. These findings indicate that employing bio-inspired optimization visibly increases the performance of CNN. The genetic algorithm achieved the best accuracy, making the model more accurate. Overall, these methods helped CNN model predict the nitrogen levels better. This methodology provides essential insights into precision farming, enabling the creation of more efficient nutrient management plans.