Identifying Incident Causal Factors to Improve Aviation Transportation Safety: Proposing a Deep Learning Approach

Table 2

A comparison of our study with extant research related to aviation reporting system.

Studies	Research target	Data set	Algorithms	Performance
Tiller et al. [14]	Analyze close call incident reports to assess severity level	117 reports from the ASRS (2014–2016)	Bliss’s taxonomy, a manual case-by-case review process	Modification on the close call taxonomy is needed, but results were not discussed quantitatively
Tanguy et al. [2]	Extract metadata and keywords from the narratives, and topic mining	86,912 qualified reports used from DGAC	N-Grams Support Vector Machine topic modeling	Incident reports classified to seven major topics, with about 78% score on average
Kuhn [15]	Automate the topic mining process	ASRS incidents from 2010 to 2015 (the exact number is not specified)	N-Grams topic modeling	Some incidents are closely related to key words, and topic modeling identified those well, but results were not evaluated quantitatively
Robinson [13]	Identify the contributing factors of the incident reports	7,484 incident reports from the ASRS	Latent semantic analysis	Identify the multiple factors of each incident; the accuracy needs significant improvement
Shi et al. [4]	Identify two primary causal factors of incidents with machine learning	168,227 incidents from the ASRS	Naive Bayes Hoeffding tree OzaBagADWIN	Automate to identify two most casual factors, and topic mining used to extract structured information
Our study	Identify the primary factor and multiple contributing factors of each incident from six most causal factors	181,651 incident reports from the ASRS	Deep recurrent neural networks	Demonstrate that deep learning is a powerful tool for processing complex textual data. We achieve best performance so far to identify the primary factor and contributing factors among related research