An active recommendation algorithm for trade union rights and interests services in power grid enterprises based on knowledge graphs and dynamic user personas

Proactive recommendation of trade union rights and interests services is essential for improving service precision and responsiveness in power grid enterprises. Existing service models often rely on manual processing or static rule matching and therefore have limited capacity to capture employees' changing needs across job roles, policy contexts, and service scenarios. To address this limitation, this study proposes an active recommendation algorithm that integrates a domain knowledge graph with dynamic user personas. The framework constructs the knowledge graph through enhanced entity recognition, relation extraction, and knowledge graph embedding, while dynamic user persona modeling captures recent changes in employee behavior. A hybrid recommendation model then combines collaborative-filtering signals, knowledge graph embeddings, and persona features to generate personalized and interpretable service recommendations. The model is evaluated using real business data from a power grid enterprise, including employee, service, complaint, safety, training, and activity-participation records. Before modeling, the data are standardized, and privacy-sensitive fields are fuzzified and de-identified. Under a unified evaluation protocol, the proposed method achieves a Precision@10 of 83.5%, a Recall@10 of 81.6%, an F1-Score of 82.5%, and a response time of 135 ms. These results indicate that integrating structured domain knowledge with dynamic user personas can improve recommendation accuracy, coverage, and interpretability in organizational service settings.