查看文章

In this work, we model the impact of weather condition on ground delay program (GDP) incidence using support vector machine (SVM) and logistic regression. We use SVM to analyze how spatial patterns of convective weather affect GDP occurrence and produce heatmaps to visualize the impact. Additionally, the SVM results are combined with local airport weather variables and airport traffic level indicator to yield a logistic model that considers both local conditions at the airport and convective weather in the surrounding area. We apply our methods to five airports: Newark Liberty International airport, John F. Kennedy International airport, LaGuardia airport, Philadelphia International airport, and Atlanta International airport. We find that the importance of convective weather depends on both its distance and direction from the airport. From the logistic regression we learn that both regional convective weather, as captured by the weights found in the SVM, and local weather are statistically significant. Convective weather is, however, the most important factor. Our models are found to have high accuracy and low false positive rates, but also low true positive rates because of the imbalance in our data.