Advanced Machine Learning Techniques for Pest Classification: A Comprehensive Review

ABSTRACT The protection of crops from pests is essential for sustainable agriculture and global food security. Traditional pest identification techniques that are based on visual examination are time‐consuming and prone to errors. Recent developments in the field of artificial intelligence (AI) and specifically deep learning (DL) enabled pests to be detected accurately and automatically. This review systematically examines DL‐based approaches, including Convolutional Neural Networks (CNNs), Transformer models, ensemble methods, and Graph Neural Networks (GNNs), for pest classification, with an emphasis on benchmark datasets, model architectures, and evaluation metrics. Transformer models, such as GNViT, achieved 99.52% accuracy and a 90.9% F1‐score on the IP102 dataset, which is approximately 10% higher than the CNNs. The Vision Transformer (ViT) model achieved 96.7% accuracy on PlantVillage. The ensemble model, like GAEnsemble, achieved excellent accuracies of 98.81% and 95.16% on D0 and SCD, respectively. CNN models had relatively lower performance on the IP102, and the GNNs showed poor performance (below 60%). This paper discusses prospective methodological enhancements, current limitations, and future prospects for developing scalable, understandable, and multi‐domain pest classification systems.