Modeling Bivariate Departure Delay Distribution for Estimating Slack Allocation

A problem in estimating slack in airline schedule connections is investigated in this paper. We propose a method for estimating slack by analyzing historical delay data and modeling the probability distribution of primary delays. Using bivariate Kernel Density Estimation, we model the distribution as a function of time and its delay duration. Then, the distribution is used to identify flights which sensitive will be delayed. The threshold technique is applied to determine the allocated slack in flight schedule connections. As a case study, we implement the proposed method to historical departure delay data collected by an Indonesian airline. According to flight delay simulation, our proposed method works well in estimating slack since it able to reduce delay propagations.