egrettably, the current infrastructure and

operating capabilities are insufficient to accommodate the

escalating demand for air travel, which is gaining popularity.

Aircraft delays have increased in frequency due to the erratic

weather patterns caused by climate change. Extensive study has

been conducted on short-term forecasting; nevertheless, due to

the intricacies of international flight networks, there is an

increasing demand for long-term flight delay prediction and

analysis. This work proposes a comprehensive technique that

encompasses data preparation, feature selection, and model

training, thereby fulfilling the criteria. The preprocessing phase

entails data cleansing and quality assurance, while the Boruta

approach is employed to identify pertinent attributes. XGBoost

enhances feature selection by retaining features that exceed a

specified relevance threshold. A hybrid model, LSTM-XGBoost,

is proposed to improve performance. It integrates the optimal

threshold with a tailored loss function. The model attained a

binary flight delay prediction accuracy of 95.57%, with a

RMSE of 0.35, a MAPE of 3.71, and a MAE of 0.2, based on

actual flight data evaluation. The results indicate that the

LSTM-XGBoost model effectively enhances long-term flight

delay predictions, hence optimising aircraft operations amid

increasing air traffic and climate-related disturbances.