Estimating wildlife strike costs at US airports: A machine learning approach

被引：10

作者：

Altringer, Levi ^{[1
,2
]}

Navin, Jordan ^{[1
]}

Begier, Michael J. ^{[3
]}

Shwiff, Stephanie A. ^{[1
]}

Anderson, Aaron ^{[1
]}

机构：

[1] USDA, Anim & Plant Hlth Inspect Serv, Wildlife Serv, Natl Wildlife Res Ctr, Ft Collins, CO 80521 USA

[2] Colorado State Univ, Dept Econ, Ft Collins, CO 80523 USA

[3] USDA, Natl Coordinator Airport Wildlife Hazards Program, Wildlife Serv, Washington, DC 20250 USA

来源：

TRANSPORTATION RESEARCH PART D-TRANSPORT AND ENVIRONMENT | 2021年 / 97卷

关键词：

Wildlife strikes; Machine learning; Cost imputation; NEURAL-NETWORK MODELS; REGRESSION;

D O I：

10.1016/j.trd.2021.102907

中图分类号：

X [环境科学、安全科学];

学科分类号：

08 ; 0830 ;

摘要：

Current lower bound estimates of the economic burden of wildlife strikes make use of mean cost assignment to impute missing values in the National Wildlife Strike Database (NWSD). The accuracy of these estimates, however, are undermined by the skewed nature of reported cost data and fail to account for differences in observed strike characteristics-e.g., type of aircraft, size of aircraft, type of damage, size of animal struck, etc. This paper makes use of modern machine learning techniques to provide a more accurate measure of the strike-related costs that accrue to the US civil aviation industry. We estimate that wildlife strikes costed the US civil aviation industry a minimum average of $54.3 million in total losses annually over the 1990-2018 period. If one assumes that wildlife strikes were underreported by as much as a factor of 3 over the same period, our estimates still fall below previous lower bound estimates.

引用

页数：16

共 24 条

[1] Development of artificial neural network models to predict driver injury severity in traffic accidents at signalized intersections [J].