Mimetic Coral Reefs Optimization (MCRO) has proven highly effective for feature selection due to its capacity to explore diverse solution spaces, enhancing model accuracy and robustness. However, integrating MCRO with local search techniques remains challenging, as it tends to be computationally intensive and prone to premature convergence. To address these issues, this paper introduces a Rank-based Adaptive brooding (RAB) mechanism, designed to refine the local mimetic search strategy within MCRO. RAB adaptively adjusts the brooding operator based on the ranks of coral larvae, minimizing disruption to high-rank larvae and harnessing the exploratory potential of lower-rank larvae. This approach promotes a more balanced exploration-exploitation trade-off, leading to faster convergence and enhanced performance in complex problem spaces. The proposed method's efficacy is tested across eight UCI datasets using KNN, Decision Tree, and SVM classifiers, and the results are evaluated by precision, recall, and F1 score. Empirical results reveal that RAB outperforms existing adaptive strategies with fixed brooding, delivering superior feature selection performance, particularly in high-dimensional datasets. Additionally, the optimization capabilities of RAB were examined using 39 CEC benchmark functions, revealing consistent improvements in feature selection accuracy while demonstrating variable outcomes in broader optimization tasks. Notably, RAB showed significant enhancements in eight benchmark cases, highlighting its potential for broader applicability in optimization scenarios.