Data Mining Of Civil Aviation Accidents Based On Multi-Association Rules

As in recent years, the proportion of civil aviation transport in whole transportation steadily improved, more and more people choose air transport as their primary way to travel. However at the same time, compared with the scale of civil aviation transportation being continuously expanded, aviation safety levels did not produce significant improvement. This thesis chooses multi-dimensional association rules as the research method, extracts historical civil aviation accident data published by authoritative survey agencies and websites, analyzes these data, researches and discovers the potential associations among them, and finally provides some suggestions and decisions-making for the further improvement in the safety of civil aviation.The concept of data mining theory and the method of association rules are discussed, the current data mining technique is clarified in several ways as well as their application in civil aviation and the whole aviation field. A definition of civil aviation security is given, the current situation of civil aviation is analysed as well as the room for improvements.It implements a horizontal comparison for data richness and data characteristics among the civil aviation accidents published by three authoritative websites, and determines the strategy of data resource and structure as: NTSB(National Transportation Safety Board) for main while ASN(Aviation Safety Network) and World Aviation Accident Database for supplement. Combined with the accident investigation reports published by authoritative survey agencies, an independent civil aviation accidents database is built which is suitable for multi-dimensional association rules analysis. Among them, the master database contains 214 global aviation accident details since 1985 while the secondary database contains 727 brief data of the global aviation accidents since 1990.According to the characteristics of multi-dimensional association rule analysis method and the data structure of civil aviation accidents, the processes and algorithms of multi-dimensional association rules are optimized, the useless information data is filtrated from the civil aviation accidents database, the data is amended and discretized, and the analysis model for civil aviation accidents is adjusted and constructed. A five-pointed star-type connection data model is determined which is centered by accident property, including aircraft property, engine property, casualty property, time property and state property.Based on the algorithm of multi-dimensional association rules applicable to the civil aviation accident information analysis, an aviation safety accident decision and analysis software is made. By using this software, several following functions can be achieved: user management, aviation accident information entry and inquiry, aviation insecurity analysis, aviation safety related site index and other functions. A userdefined queries and unsafe aviation accident causation analysis are achieved, providing an effective research method and decision support in order to improve civil aviation safety.