—Most current research applying data mining
techniques to aviation safety focus on the acquisition and
selection of data attribute fields, while the selection and analysis
of classification algorithms has been largely overlooked. This
study seeks to address this gap as a means of identifying key
factors which result in aviation fatalities. The study data is
taken from the Accident/Incident Data System (AIDS) flight
accident database, which comprises an imbalanced dataset. The
Synthetic Minority Oversampling Technique (SMOTE) is used
to minimize the impact of data imbalance on classification
results for categories with small datasets. Feature attribute
selection was implemented through information gain (IG). This
study also develops an Associative Petri Net (APN) model for
comparison against five other classification methods: which are
Naive Bayes, BayesNet, Support Vector Machine (SVM),
Decision Tree (C4.5), and Radial Basis Function Network
(RBFN). IG is used to identify 10 important factors: Event City,
Event State, Flight Phase, Aircraft Model, Aircraft Series,
Operator, Primary Flight Type, Flight Conduct Code, Flight
Plan Filed Code, and Nbr of Engines. Results show that our
proposed APN model had the highest overall accuracy rate,
F-measure score, G-mean score among all algorithms. APN is
based on the Apriori algorithm to provide a rule-based concept.
Thus using the APN algorithm to build the model could produce
an expert system for flight accident prediction.
—Aviation safety, attribute selection,
imbalanced dataset, SMOTE, associative petri net.
H. S. Chiang, T. C. Hsieh, and M. Y. Chen are with the National Taichung
University of Science and Technology, Taichung, 40444 Taiwan (e-mail:
firstname.lastname@example.org, email@example.com, firstname.lastname@example.org).
C. C. Chen is with National Chung Hsing University, Taichung, 402
Taiwan (e-mail: email@example.com).
Neil Y. Yen is with University of Aizu, Aizu-Wakamatsu, Fukushima
Pref. 965-8580 Japan (e-mail: firstname.lastname@example.org).
Cite: Hsiu-Sen Chiang, Tsung-Che Hsieh, Chia-Chen Chen, Neil Y. Yen, and Mu-Yen Chen, "Using Associative Petri Net with Over-Sampling Techniques to Construct an Aviation Incident Prediction Model," International Journal of Future Computer and Communication vol. 6, no. 1, pp. 21-26, 2017.