In light of the rapidly growing passenger and flight volumes, airports seek for sustainable solutions to improve passengers’ experience and comfort, while maximizing their profits. A major technological solution towards improving service quality and management processes in airports comprises Internet of Things (IoT) systems that realize the concept of smart airports and offer interconnection potential with other public infrastructures and utilities of smart cities. In order to deliver smart airport services, real-time flight delay data and forecasts are a critical source of information. This paper introduces an essential methodology using machine learning techniques on Apache Spark, a cloud computing framework, with Apache MLlib, a machine learning library to develop and implement prediction models for air flight delays that could be integrated with information systems in order to provide up-to-date analytics. The experimental results have been implemented with various algorithms in terms of classification as well as regression, thus manifesting the potential of the proposed framework.