Abstract:

Flight delays are constantly rising, causing airlines to face greater financial issues and customer dislike. To address this issue of flight delays, supervised machine learning models were deployed. For the forecast, a data collection that captures information about planes departing from JFK airport over the course of a year was employed.To estimate airline arrival delays, the prediction model described in this research employs supervised machine learning approaches. According to the Federal Aviation Administration (NASEM, 2014), delayed flights are those that are delayed for more than 15 minutes over their scheduled time. When a flight is delayed, the airlines and passengers are usually the ones that suffer the most. The delay of one flight may spread and impact subsequent flights. Increased delays lower passenger demand for airlines. Furthermore, airfares are higher on routes with a higher number of delays. As a consequence, buffer time is added to timetables to prevent delays, and flights can still arrive on time. However, this scenario is less likely to occur in congested or bigger airports. With a longer buffer time, fewer flights are booked for the day (NEXTOR, 2010). Domestic flight data and meteorological data from the United States were gathered and used to train the prediction model from July to December 2019. The Logistic Regression, SVM, Decision Tree, and Random Forest approaches were taught and evaluated in order to complete the binary categorization of flight delays. The algorithms were compared using four metrics: accuracy, precision, recall, and f1-score. Flight and meteorological data were fed into the model.

Description:

Flight, delays,Aviation, Random Forest, train, prediction, DT, RF