The code of Urban Building Clustering via Topological Manifold and Cross-View Diversity Fusion (TMCD).

Accurate clustering of urban buildings is essential for uncovering urban patterns. However, existing building clustering methods still encounter challenges in similarity measurement and handling multi-view structural inconsistency within complex spatial distributions. Specifically, Euclidean-based metrics often fail to capture intricate topological and geometric relationships among buildings, while current multi-view clustering methods tend to overlook structural inconsistencies between spatial and non-spatial views, limiting robustness. To address these challenges, this study proposes TMCD, a novel multi-view clustering framework that integrates topological manifold modeling with cross-view diversity detection. First, TMCD constructs three complementary views (centroid distance, outline distance, and non-spatial attributes) and learns latent manifold topology structure between buildings in each view. Then, a cross-view diversity detection mechanism identifies inconsistencies, producing a pure graph for each view, which is then fused into a consensus graph with a clear clustering structure. Finally, an alternating iterative optimization algorithm jointly learns the topological correlations, multi-view consistency, and consensus structure within a unified framework. Extensive experiments on 15 urban community building datasets from Wuhan and Chengdu show that TMCD outperforms 12 baseline methods, improving Adjusted Rand Index (ARI) by up to 32% and Accuracy (ACC)/F-score by over 20%, with strong adaptability to both complex and linearly arranged urban patterns.