Automatic generation of monitoring report based on large language model and knowledge graph inference
Zengxiong Chen, Yanfeng Qiu, Longlong Yang, Baijian Liao, Defa Cao · Results in Engineering · 2025
• The method integrates LLMs and knowledge graphs, achieving > 86 % efficiency gains in data processing. • It employs ant colony algorithms for multi-modal data integration and MFCC-GPT fusion for precise feature extraction. • Results show 0.09 % packet loss, > 90 % knowledge graph coverage, and 4.29 μs extraction time for 10 data points. • Future work includes expanding knowledge graph content and improving model interpretability for broader applicability. • Enhanced hardware compatibility and deep learning integration are prioritized for complex monitoring scenarios. To optimize traditional data collection, sorting, analysis, and report preparation processes, this paper introduces an automatic monitoring report generation method grounded in large language models and knowledge graph inference. Leveraging stream processing technology for data loading, transformation, and extraction, and the ant colony algorithm for querying and integrating environmental, equipment, and virtual monitoring data, this method enhances feature extraction and data using Mel-frequency Cepstral Coefficients (MFCC) for speech signals and Generative Pre-trained Transformer (GPT) series models. Text-based semantic similarity evaluation and a replication mechanism in time knowledge graph inference further refine the process. The experimental results show significant improvement: for 300,000 data points, the extraction time of the design method is 39.2 s, and the efficiency is improved by 86.3 %, 87.6 %, 88.5 % and 85.5 %, respectively, compared with the methods in [ [3] , [4] , [5] , [6] ]. For 10 data points, the extraction time was only 4.29 μ s. With 180 nodes, the packet loss rate is a mere 0.09 %, and the knowledge graph coverage exceeds 90 %. This method's unique contribution lies in its seamless integration of advanced technologies, ensuring high efficiency and accuracy in data processing and analysis.