An AI-based Model for Predicting Flight Delays to Enhance Air Traffic Operations

Abstract

Flight delays are a global challenge with significant economic and operational impacts. This study aims to develop a predictive model based on machine learning algorithms to enhance the efficiency of air traffic management. To achieve this, the study adopts an analytical and applied approach, utilizing a comprehensive historical dataset from the U.S. Bureau of Transportation Statistics (BTS), which includes over 5.8 million domestic flights in 2015. The research methodology entailed precise data processing steps, including cleaning, feature engineering, and transforming the prediction task into a binary classification problem. The model was constructed using the Random Forest algorithm, and its performance was optimized through the GridSearchCV technique to select the best parameters. To further increase the model's efficiency, the Recursive Feature Elimination (RFE) method was employed to identify the 20 most influential features for prediction. The study yielded highly significant results, with the proposed model achieving a high predictive accuracy of 98.41% in determining whether a flight would be delayed or not. Feature importance analysis revealed that factors such as "Departure Delay," "Scheduled Duration," and "Month" were the most influential in the prediction. These findings demonstrate the model's effectiveness in providing valuable insights that can be leveraged for proactive decision-making. Based on these results, the study recommends integrating the predictive model into current air traffic management systems to improve operational planning and mitigate losses. It also proposes that future research should extend the scope to include predicting the actual delay duration and exploring the model's application on global data to assess its generalizability.