In order to make the historical accident experience play an efficient and accurate support role in emergency decision-making of subway construction collapse accidents,ageneration method of emergency measures for subway construction collapse accidents based on case-based decision theory(CBDT) and improved K-nearest neighbor(KNN) algorithm was proposed.From the perspective of CBDT,based on Chinese word segmentation technology,the accident description unit factors were screened out to structurally record accident information,then historical accident case data base was build.Based on the improved KNN algorithm,a five-step accident similarity calculation algorithm was designed,including weight,structural similarity,local similarity,substitutability and global similarity calculation,to retrieve similar historical accidents and output suggestions on emergency measures.Finally,the feasibility and effectiveness of the method proposed in this study were verified by case study.