Construction of traffic accident knowledge graph based on the correlation analysis of risk factors

With an increase of vehicles, traffic accidents have also risen. This paper presents a novel approach to create a traffic knowledge graph using a keyword extraction algorithm to analyse accident data from a specific city, focusing on identifying key terms related to the causes of accidents. The data are analysed from four aspects: human factors, vehicles, road conditions and environmental factors, to construct the knowledge graph. TextRank is a graph-based unsupervised keyword extraction method that ranks words based on their cooccurrence in a sliding window. The findings indicate that the improved TextRank algorithm, which incorporates word vectors and a multi-feature weighting mechanism, outperforms traditional TextRank and inverse document frequency methods in keyword extraction. The present TextRank algorithm effectively combines word-specific attributes and structural features, delivering better extraction performance.