Association rule mining of aircraft event causes based on the Apriori algorithm

Area of Science:

• Aviation Safety
• Data Mining
• Accident Causation Analysis

Background:

• Understanding complex aircraft events requires analyzing causative factors.
• Traditional statistical methods can alter interrelations among causes when handling missing data.
• Association rule mining offers a robust approach to uncover hidden patterns in aviation incident data.

Purpose of the Study:

• To mine association rules between trigger probability and relative strength in aircraft events.
• To identify causative item sets and establish code indicators based on the accident causation 2-4 model.
• To enhance the Apriori algorithm for analyzing aviation incident data without altering causal interrelations.

Main Methods:

• Data preprocessing using clustering and TF-IDF value calculation.
• Modified Apriori algorithm incorporating frequent item consideration to avoid NaN value issues.
• Network graph visualization to represent association rules between unsafe events and causes.

Main Results:

• Identified specific factors highly correlated with general unsafe events (e.g., improper energy allocation, controller oversight).
• Found notable correlations for general incidents (e.g., failure to rectify incorrect recitation).
• Determined strong correlations for serious incidents (e.g., inadequate manual promotion, insufficient safety management).

Conclusions:

• The study quantitatively reveals patterns and characteristics of interactions among historical aircraft events.
• Findings highlight directions for controllable prevention and prediction of aircraft events.
• The enhanced Apriori algorithm provides a valuable tool for aviation safety analysis.