Hybrid Optimization of K-Means Clustering for BIG Data Using Deep Reinforcement Learning and Quantum Whale Optimization Algorithm

In this paper, a novel hybrid clustering framework, Deep Reinforcement Learning–Quantum Whale Optimization–KMeans (DRL–QWOA–KMeans), is proposed to overcome major limitations of conventional K‑Means, including initialization sensitivity, premature convergence, and reduced clustering accuracy on complex data distributions. In the proposed model, Deep Reinforcement Learning (DRL) is employed to dynamically balance the exploration–exploitation process within the Quantum Whale Optimization Algorithm (QWOA), which adaptively optimizes the initial centroids of K‑Means to achieve global convergence. Extensive experiments conducted on benchmark UCI datasets demonstrate the robustness and efficiency of the proposed method compared with classical meta‑heuristic‑based clustering algorithms such as GA, PSO, DE, WOA, and QWOA. Quantitative evaluation using multiple performance metrics—WCSS, AAcc, ASen, ASpe, and FscoreM—indicates a significant improvement in intra‑cluster compactness and inter‑cluster separability. Moreover, visual analysis confirms that DRL–QWOA–KMeans yields more stable and interpretable clusters compared to standard K‑Means. These findings highlight the competitive potential of DRL‑guided quantum optimization for complex clustering problems and provide a foundation for future investigations on larger-scale or streaming data environments