<p>In the era of big data, identifying and optimizing data risk patterns is crucial for enhancing decision-making accuracy and system security. However, existing methods often treat all features equally or lack sensitivity to varying risk levels, which limits their effectiveness in complex risk environments. This paper proposes AWK-RPO (Adaptive Weighted K-means for Risk Pattern Optimization), a novel clustering-based framework that integrates adaptive feature weighting with risk-aware optimization. Unlike traditional K-means algorithms that treat all features equally, AWK-RPO iteratively adjusts the weight of each feature based on its contribution to the overall risk, enabling the algorithm to highlight critical risk-related attributes while suppressing noise. Furthermore, a risk-sensitive distance metric is employed to improve the clustering accuracy for high-risk data points. The proposed method incorporates a feedback loop that refines the feature weights and cluster boundaries until convergence, ensuring both robust pattern recognition and optimized risk distribution. Experimental results on multiple real-world datasets demonstrate that AWK-RPO outperforms existing K-means variants in terms of clustering quality, risk detection rate, and optimization efficiency.</p>