Age-related Macular Degeneration (AMD) is a leading cause of visual impairment globally, with neovascular AMD (nAMD) posing significant challenges for effective treatment. This study presents a novel approach leveraging the Clustering-constrained Attention Multiple instance learning Single-branch (CLAM-SB) framework combined with ConvNeXt V2 for feature extraction to predict neovascular AMD (nAMD) progression from optical coherence tomography (OCT) images. Our method addresses the limitations of traditional instance-based approaches by aggregating information across multiple images, resulting in enhanced prediction stability and accuracy. Additionally, we mitigate class imbalance using a weighted random sampler, further improving model performance. The proposed approach not only demonstrates superior efficacy in monitoring disease activity but also secured first place in the preliminary round of the Monitoring Age-related Macular Degeneration Progression in Optical Coherence Tomography Task 2 at MICCAI 2024, underscoring its potential as a robust tool for individualized treatment planning in patients undergoing anti-VEGF therapy.