Mining for Information in Accident Data
Abstract
Purpose. This project evaluated WinMine, an analytic tool developed by Chickering, Heckerman, Meek, Platt, and Thiesson (2000) to determine its usefulness for identifying higher-order relationships in research data from dynamic and high-consequence aviation events. Traditionally, researchers have relied on several types of analyses to better understand the relationships between factors related to an outcome. However, researchers need an analytic approach that can clearly illustrate the interactions among causal factors as probabilities associated with the chain of events. Method. A convenience sample of aviation accident data previously classified using the Human Factors Analysis and Classification System (HFACS; Shappell & Wiegmann, 2000; 2001) was used to evaluate WinMine in contrast to traditional methods, such as bar graphs, contingency tables, and odds ratios. Results. WinMine showed an advantage when compared with other methods because it graphs quantifiable interrelationships between factors and illuminates the underlying hierarchical structure of variables. Discussion. Each technique examined contributed toward understanding the causal factors; however, WinMine provided a better picture of the factor interrelationships than the other methods.
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 01, 2006
- Accession Number
- ADA464086
Entities
People
- Julia Pounds
- Lena Dobbins
- Nelda J. Milburn
- Scott Goldman
Organizations
- Federal Aviation Administration