Reduce unexpected airline diverts: modelling with time series analysis and neural network

In this study, a decision support system is designed in order to minimize the number of flights that are diverted unexpectedly. The aim is to reduce the expenses that arise when the aircraft is not able to land on the targeted airport due to the unfavorable weather conditions, such as rescheduling the timetable, overuse of aircraft fuel than planned, passengers' accommodation and ticket reissue. In order to reduce such temporal and financial losses caused by diverted flights, decision to take off or not is made before departure, while the decision to land or not is made during flight, after a brief analysis based on weather data of target airport. For the aircraft to land on target airport as scheduled, it is crucial that the weather forecasts for visibility range, ceiling and wind speed are within the limits of the safe flight requirements. Considering the significance of this decision regarding by finance, there is a need for a decision support system that is capable of boosting the process through optimal decision-making by forecasting airport weather conditions. In the first part of the study, weather is forecast using regression and time series analysis, of which methods can be detailed as auto regressive (AR), moving average (MA), auto regressive integrated moving average (ARIMA) and vector auto regressive (VAR). Although such forecast methods are relatively effective in achieving the desired result, neural network and fuzzy logic techniques are expected to present more accurate forecast with their complicated and advanced algorithm structure. In the second part of the study, neural networks are created with using MATLAB. The results which is obtained with these methods are compared time series analysis results. Improvement is measured by accuracy of the decisions of diverted flights. The measurements are recorded on the confusion matrix. Key Words: Divert, Weather Forecast, Regression, Time Series, Decision Support System, Neural Network, RNN, LSTM .