This paper proposes a fuzzy matching-based algorithm for Internet of Things (IoT) public information retrieval to address challenges like multi-source heterogeneity and semantic ambiguity. Our method utilizes multi-round interactive feedback to analyze user search history, incorporates role-based segmentation, and extracts high-frequency keywords through semantic preprocessing. Normalized weights of Document Term Frequency (DTF), Document Frequency (DF), and Inverse Document Frequency (IDF) are computed and fuzzified to construct a feature library. An inverted index with bitwise operations enables efficient matching, while subspace fuzzy clustering and integer linear programming optimize semantic relevance. Experiments on 10,000 records show superior Normalized Discounted Cumulative Gain (NDCG) (peak value of 110) and user engagement depth (peak 600) compared to benchmarks, with stable performance (fluctuation range: 7.77). The algorithm proves robust for complex IoT scenarios.