Applying Large Language Models and Knowledge Graphs for Intelligent Mobile Network Complaint Processing

This study examines the application of generative AI in mobile network operations, focusing on customer complaint classification. By leveraging natural language processing, complaints are summarized and categorized to assist service personnel in accelerating root cause analysis. This facilitates targeted interventions to improve service quality and reduce recurring issues. We introduce a novel prompt strategy that integrates Chain-of-Thought (CoT) reasoning with knowledge graphbased retrieval, enabling guided multi-hop reasoning within a large language model (LLM). To evaluate its effectiveness, we compare our approach against baseline prompting methods, including Zero Shot, Few Shot, and knowledge graph-guided strategies. Experimental results show that our method achieves superior classification accuracy. Our findings also indicate that generative AI can significantly reduce the manual workload in complaint handling, cutting analysis and processing time per case by approximately $40 \%$. These findings highlight the potential of combining generative AI with structured domain knowledge to significantly enhance the efficiency and accuracy of mobile network maintenance operations.