Association Rule Mining of Aircraft Incident Causes based on Apriori Algorithm

Abstract

Abstract Combining accident causation 2–4 model with Apriori algorithm of data mining technology, an association rule of aircraft incident causation is proposed. The data cleaning model based on expert system is adopted on the python platform with the statistics of unsafe events in Civil Aviation of China in 2023. The accident cause 2–4 model divides the factors leading to aircraft unsafe incidents into internal and external factors of HMI organization. The top 30 TF-IDF words of each level are taken as the cause factors of the level, and strong association rules between aircraft unsafe events and cause factors are found through the iterative method of searching layer by layer with Apriori algorithm. Taking the 3-year unsafe incidents of Southwest Air Traffic Management Bureau in CAAC as an example, 90 cause-causing factor sets were identified, and unsupervised and supervised learning methods were respectively adopted. The support, confidence and improvement degrees of the two methods were compared using the complex network diagram of aircraft accident cause-causing generated by Gephi. The comparison results revealed that the analysis of association rules among factors is no longer limited to a single node, the index difference of association rules is increased, and the strong causality is more significant. It points out the direction of accident prevention and prediction.