The processing of imbalanced data in machine learning is a significantly challenging situation. Imbalanced data impacts the performance of highly stable, strong, and reliable models. In sensitive domains like healthcare, fault prediction, cyber security, etc., if a model is trained using imbalanced data, it can lead to severe consequences. To address this imbalance issue, this paper offers a novel approach based on the Ordering Points To Identify the Clustering Structure (OPTICS) algorithm called an Adaptive OPTICS (A-OPTICS). The approach integrates density-based adaptive sampling, cluster-aware feature engineering, and ensemble learning strategy. The conventional approach tackles class imbalance uniformly, but this approach concentrates on the local density information and cluster structures to improve the learning process and maintain model generalizations. It prioritizes minority instances and low-density locations based on reachability distances, which guarantees meaningful representation. The Cluster-aware feature engineering focused on local density patterns which provide the ability to distinguish between minority and majority classes effectively. A-OPTICS offers an effective and viable solution to the prevalent problem of imbalanced data by providing a fair learning process. The experimental results on various benchmark imbalanced datasets demonstrate significant classification performance compared to the existing state-of-the-art methods. This illustrates its usefulness and robustness across several domains. Therefore, this proposed approach must be a feasible option for real-world applications.