Extreme gradient boosting approach for air traffic data analysis in pandemic conditions

World Health Organization has described COVID-19 as a serious disease that spreads quickly from person to person through various contact routes and respiratory droplets. Airplanes are a trendy mode of transportation for everyday travel and tourism. An airport is a complex system that is an essential link in aviation. Aviation is one of the business sectors significantly affected by the COVID-19 outbreak, which had a global impact. Various countries established travel restrictions. Accurate analysis of air traffic plays an important role in enhancing the performance of the air transport network (ATN). Numerous methods have been developed for analyzing big data on air traffic during the COVID-19 pandemic. However, accuracy improvements were not achieved, and time was not reduced. To address this, a novel technique called the Stochastic Quantile Extreme Gradient Boost (SQEGB) is introduced. The SQEGB technique includes feature selection and classification. First, the number of data samples is collected from the airport traffic datasets. Following data collection, Sokal Sneath Stochastic Neighbor Embedded feature selection is performed within the SQEGB technique to select the most significant features for air traffic data analysis with less time and space. Once the relevant features have been identified, the next step involves classifying the data using the Quantile Regressive Extreme Gradient Boost classifier. The designed classifier categorizes the data based on the number of trips a flight can take in a day compared to the average number of trips on the same day. If the traffic (i.e., the number of flight trips) is lower, the risk of disease spread is significantly reduced with higher accuracy. Consequently, this approach enables accurate analysis of air traffic data. Experimental evaluation is estimated using various quantitative metrics, including air traffic prediction accuracy, mean absolute prediction error, and air traffic prediction time and space complexity. The results of the proposed SQEGB method achieve 7% prediction accuracy, 64% of minimal mean absolute prediction error, 12% and 13% of reduced air traffic prediction time and space complexity when compared to existing methods.