부산대학교 도서관

This report describes initial experimentation to understand and determine the feasibility of developing a machine-learning based approach to forecast stratus clearing times at San Francisco International Airport (SFO). Marine stratus conditions along the approach path into SFO airport frequently require the issuance of a Ground Delay Program by the FAA. To minimize the cost and delay impacts of the reduced arrival capacity, it is of interest to predict, well in advance, when these stratus events will clear. This prediction of the arrival capacity increase permits planning an optimal release schedule for ground-delayed aircraft, such that aircraft arrive soon after the stratus has cleared, without affecting the safety of landing aircraft. Two different machine learning approaches have been developed and are described in this paper, including machine learning training and testing results.