With the advancement of internet technology, diverse networks have emerged. Analyzing relationships between network entities can enhance our understanding of real-world network evolution, where link prediction is crucial. However, current link prediction methods face two key challenges. First, most algorithms extract node features from local subgraphs, overlooking global characteristics. Second, they struggle to adapt to graph structures with widely varying node counts. To address these issues, we propose HSHLP (High Structural Hole Link Prediction), a novel approach that leverages structural hole features across multi-order paths. By incorporating structural hole theory, HSHLP uncovers hidden node-linking patterns and builds effective feature representations through multi-order neighborhood extraction. Experiments show that HSHLP not only increases prediction accuracy but also remains robust across networks of varying scales.