Knowledge-driven intelligent recommendation method for emergency plans in water diversion projects

Abstract

Abstract The emergency plans for water diversion projects suffer from weak knowledge correlation, inadequate timeliness, and insufficient support for intelligent decision-making. This study incorporates knowledge graph technology to enable intelligent recommendations for emergency plans in water diversion projects. By employing pre-trained language models (PTMs) with entity masking, the model's ability to recognize domain-specific entities is enhanced. By leveraging matrix-based two-dimensional transformations and feature recombination, an interactive convolutional neural network (ICNN) is constructed to enhance the processing capability of complex relationships. By integrating PTM with ICNN, a PTM–ICNN method for joint extraction of emergency entity relationships is constructed. By utilizing the Neo4j graph database to store emergency entity relationships, an emergency knowledge graph is constructed. By employing the mutual information criterion, intelligent retrieval and recommendation of emergency plans are achieved. The results demonstrate that the proposed approach achieves high extraction accuracy (F1 score of 91.33%) and provides reliable recommendations for emergency plans. This study can significantly enhance the level of intelligent emergency management in water diversion projects, thereby mitigating the impact of unforeseen events on engineering safety.