Publication

Abstract : Age-related macular degeneration (AMD) is a degenerative retinal condition and a primary cause of permanent visual impairment in the elderly. The precise categorisation of AMD stages is essential for accurate screening, prompt intervention, and tracking of disease progression. This paper introduces a Light Gradient Boosting Machine (LightGBM) model for the multiclass classification of AMD stages using the publicly accessible AMDP dataset from Kaggle. The collection comprises retinal images categorized into various disease stages, from normal to severe AMD. The images were pre-processed and converted into a structured dataset appropriate for model training. LightGBM was chosen for its efficiency, scalability, and robust performance in multiclass classification problems. The model underwent training and evaluation using a stratified train-test split, guaranteeing balanced representation throughout all stages of AMD. Experimental findings indicate that the proposed LightGBM methodology achieved an overall classification accuracy of 99.96%, proficiently differentiating among normal, early, intermediate, and advanced stages of AMD. These findings validate the efficacy of gradient boosting–based techniques as a dependable and interpretable approach for classifying stages of AMD. The research highlights the practicality of using Ligh