Abstract
<title>Abstract</title> <p>This paper introduces a novel Chaotic Improved Cloud Drift Optimization (CI-CDO) algorithm that enhances the original Cloud Drift Optimization (CDO) through a physically grounded atmospheric model driven by chaotic dynamics. The proposed framework redefines the search process using three complementary mechanisms: a Cloud Splitting Mechanism for global exploration, a Cloud Merging Mechanism for local exploitation, and an adaptive Atmospheric Turbulence mechanism governed by the Atmospheric Turbulence Index (np) to dynamically regulate the transition between them. Furthermore, a logistic chaotic map, adaptive drift coefficients, and a structured multi-candidate update strategy improve population diversity, strengthen convergence behavior, and reduce premature stagnation in complex optimization landscapes. To evaluate its effectiveness, CI-CDO was comprehensively assessed through three experimental scenarios. First, it was benchmarked on the CEC 2021 test suite, where it achieved superior or highly competitive performance across unimodal, multimodal, hybrid, and composition functions compared with state-of-the-art metaheuristic algorithms. Second, CI-CDO was adapted for binary feature selection on 21 UCI datasets using a wrapper-based *k*-Nearest Neighbors classifier, consistently achieving high classification accuracy while selecting compact feature subsets. Finally, the proposed algorithm was applied to real-world seasonal water quality and temperature forecasting using data collected from Egyptian monitoring stations, demonstrating strong predictive performance, robustness, and generalization capability. Overall, the proposed atmospheric mechanisms and chaos-driven search strategy establish CI-CDO as a robust and physically inspired optimization framework for solving both benchmark optimization problems and real-world engineering and data science applications.</p>