<p>Graph-based multi-view semi-supervised learning has emerged as a prominent research area in recent years because it uses the graph structure of the data, which helps to better capture the relationships between samples. However, existing methods have two main limitations: (1) They generally involve two independent steps: building a similarity graph and spreading labels from labeled to unlabeled data. But the two-step method often has suboptimal results. (2) They only emphasize inter-view compatibility but neglect the differences within intra-view sub-features. To overcome the two limitations, we present a novel graph-based multi-view semi-supervised classification method, named Semi-Supervised Classification based on Adaptive Weighting of Inter-view and Intra-view Sub-features (SSC_AWIIS). Specifically, our proposed method unifies label propagation and similarity graph construction within a single framework, which reduces reliance on the accuracy of the generated similarity graphs. And it takes into account both inter-view compatibility and intra-view sub-feature heterogeneity. Moreover, our proposed method can automatically assign optimal weights to each view and sub-feature without explicit weight definitions. Finally, we conduct the comparison experiments on multiple multi-view datasets to verify the efficacy of our proposed SSC_AWIIS.</p>