Selection of Best Machine Learning Model to Predict Delay in Passenger Airlines

Abstract

Over the past years, flight delay has been a critical concern in the aviation sector due to the increased air traffic congestion worldwide. Moreover, it also prolongs the other flights, which can discourage users from traveling with the particular airline. As a result, we proposed a model to predict the overall flight delay using a random forest and path-finding algorithm. The proposed model focuses on searching flights (can be nonstop or connecting) between the source and destination at the earliest. The proposed model identifies the fastest flights between source and destination based on the input by the user using some open source/public Application Programming Interface (APIs), which are further inserted into Neo4j to convert it into a JavaScript Object Notation (JSON) format. Finally, the experimental results on the real-time data set show the proposed model’s effectiveness compared to the state-of-the-art models. The results and analysis yield an accuracy of 98.2% for delay prediction on historical data using the random forest algorithm.

Keywords

• Flight delay
• flight search
• Neo4j
• python
• random forest