Development of Predictive AI for Flight Delays using Machine Learning
Loading...
Date
2024
Journal Title
Journal ISSN
Volume Title
Publisher
UMT, Lahore
Abstract
Flight delays pose a significant challenge for the aviation industry, impacting airlines, airports, air
traffic controllers, facility managers, and passengers. The development of a highly accurate
prediction model is the need of the hour to enable informed decision-making. This paper proposes
a model based on flight data attributes and adverse weather conditions, utilizing data mining and
supervised machine learning algorithms. Deep machine learning algorithms and neural network
models were applied to make accurate predictions. Each model's prediction accuracy and receiver
operating characteristic (ROC) curves were compared to evaluate their performance. This model
was developed in Pakistan at three major airports, including Islamabad (ISB), Lahore (LHE), and
Karachi (KHI). The ultimate objective of the project is to address the significant shortcomings in
the region's capability to predict aircraft delays with precision. This study is competitive as
previous models cover one airport, but it covers three. For the prediction phase, flight attributes
and weather data were input into the model. The trained model will then predict whether a flight
will be on time or delayed based on specific groupings. The model achieved a mean absolute error
of 23, which means the accuracy rate was remarkable, given the specific key attributes of data and
weather. This proposed model aims to boost operational efficiency and passenger satisfaction by
further enhancement.