Employing hybrid GPR-KNN with diverse acquisition functions for the prediction of current consumption in Aeronautical Ground Lighting systems

Numerous studies have explored runway visibility prediction. However, relatively few have focused on predicting nominal current output based on visibility, time, and acquisition functions, as effective management of aeronautical lighting is crucial for optimizing energy consumption and reducing environmental impact. This study decisively targets optimizing nominal current output of runway edge lights at Subang Airport using advanced Machine Learning (ML) techniques, leveraging a range of acquisition functions and comprehensive meteorological data. A hybrid combination of Gaussian Process Regression, which is particularly effective in accounting for data-specific uncertainties and non-linear relationships, and Nearest Neighbor Classifiers are utilized to accurately predict the nominal current output based on predicted visibility and time. Both 5-fold crossvalidation and holdout validation were performed to ensure robust evaluation. Model performance is assessed using key metrics such as accuracy, precision, recall, and F1 score. The findings explicitly demonstrate that the Expected Improvement (EI) acquisition function outperformed others, which is the most accurate in predicting the nominal current output in both validation methods by achieving 99.62% accuracy. In conclusion, this study presents a groundbreaking approach to predicting and improving nominal output current for runway edge lights by applying the EI acquisition function.