To conduct a deep-level analysis of the causes of civil aviation accidents for the purpose of preventing accidents, addressing the current incompleteness of accident data statistics, and ensuring the credibility and depth of accident causation analysis, accident report data from global civil aviation accidents in the year 2000 were collected as samples and subjected to structured processing. The accident data were categorized based on stages, types, and other fields, and statistical analysis was conducted. Utilizing the K-means algorithm, civil aviation safety risk factors were clustered from multiple perspectives, comparing the differences between structured and unstructured processing results. Various risk factors' combined impact on accidents was comprehensively considered. Key risky factors from the word cloud diagram and latent topic distribution in text data identified through LDA (latent dirichlet allocation) topic modeling were used to analyze accident causation. This study provides valuable reference information and safety recommendations for future civil aviation safety from multiple aspects.