Prediction of Gate to Gate block time for scheduled flights is considered as one of the challenging tasks in Air Traffic Flow Management (ATFM)system. Establishing an effective and practically reliable model to manage the problem of block time variation is a significant work. The airlines do tend to pad or inflate block time to Actual Block time to calculate Schedule block times which is approved by aviation regulator. This will lead to flaws in air traffic flow strategic decision-making and in turn affect the efficiency, estimation and undesirable delays, which leads to traffic congestion and inefficient ground delay programs. This study evaluates the effectiveness of nonlinear and time varying regression models to predict block time with minimal attributes in order to solve the problem of difficulty in predicting the block time variation. The key research outcome of this paper is to trace the temporal variations of flying time for different aircraft types and to predict the variation of actual arrival time from the scheduled arrival time at the destination airport. Ultimately, a combination of M5P regression model and logistic regression model is proposed to predict early, delayed and on-time conformity with approved schedules. Analysis based on a realistic data set of a domestic airport pair (Mumbai International Airport and New Delhi International Airport) in India shows that the proposed model is able to predict in block time at the time of departure with an accuracy of minutes for of test instances. As a result of the scheduled arrival time performance (early, delayed and timely) has been classified accurately using Logistic regression Classifier of machine learning. The test results show that the proposed model uses a minimum number of attributes and less computational time to more accurately predict the actual arrival time and scheduled arrival performance without details on the weather.


To Airports Authority of India for providing study data