Sensitive Information Filtering, Computational Framework, Large-Scale Social Media Streams, Topic Similarity Clustering, Real-Time Data Processing, Large-Scale Data Engineering Abstract: The exponential growth of user-generated content on social media platforms has introduced significant challenges in automatically detecting and filtering sensitive information in large-scale data streams. Existing methods often suffer from low adaptability, incomplete keyword coverage, and inefficient filtering performance in computational engineering scenarios. This paper presents a novel computational framework for sensitive information filtering, integrating topic similarity clustering and KNN-based classification to address large-scale social media processing demands. The framework establishes a dynamic feature lexicon through semantic clustering of related terms, enabling adaptive keyword expansion and updates to enhance filtering comprehensiveness. Experimental results on real-world social media datasets demonstrate that the proposed framework achieves a filtering accuracy exceeding 90%, significantly outperforming traditional methods. It also reduces runtime latency and improves scalability, making it suitable for real-time processing of large-scale information streams. This study provides an effective technical solution for enhancing the accuracy, efficiency, and robustness of sensitive information filtering in computational engineering contexts involving massive social media data. https://mc.manuscriptcentral.com/jcmse Journal of Computational Methods in Science and Engineering For Peer Review A Computational Framework for Sensitive Information Filtering on Large-Scale Social Media Streams ABSTRACT The exponential growth of user-generated content on social media platforms has introduced significant challenges in automatically detecting and filtering sensitive information in large-scale data streams. Existing methods often suffer from low adaptability, incomplete keyword coverage, and inefficient filtering performance in computational engineering scenarios. This paper presents a novel computational framework for sensitive information filtering, integrating topic similarity clustering and KNN-based classification to address large-scale social media processing demands. The framework establishes a dynamic feature lexicon through semantic clustering of related terms, enabling adaptive keyword expansion and updates to enhance filtering comprehensiveness. Experimental results on real-world social media datasets demonstrate that the proposed framework achieves a filtering accuracy exceeding 90%, significantly outperforming traditional methods. It also reduces runtime latency and improves scalability, making it suitable for real-time processing of large-scale information streams. This study provides an effective technical solution for enhancing the accuracy, efficiency, and robustness of sensitive information filtering in computational engineering contexts involving massive social media data. Keywords: Sensitive Information Filtering; Computational Framework; Large-Scale Social Media Streams; Topic Similarity Clustering; Real-Time Data Processing; Large-Scale Data Engineering. I.