ScholarWorks@경상국립대학교

초록

With the increasing demand for aviation, interest in machine learning is growing as a way to analyze airplane accidents for aviation safety. While the existing researches mainly have interest in constructing a classification model of casualty grades in airplane accidents and evaluating its predictive performance, the purpose of this paper is to pursue a prediction model considering data imbalance and the creation of causal rules between accident factors and casualty grades in airplane accident. Accident factors can be evaluated through feature importance of casualty grades classification model in airplane accident. In particular, casualty accidents can be more effectively classified through the gradient boosting classifier model under the data imbalance existing in airplane accident data and its performance is compared with that of the random forest model. And using accurately classified data, judgement rules for predicting casualty grades of airplane accident can be constructed using a decision tree. This paper can contribute to more precise prediction or judgment with higher recall classification and providing rules in identifying casualty grades resulted from airplane accidents.

키워드