Abstract
Load Factor (LF) calculates the sold-to-available seats ratio and defines the plenitude of commercial flights. It is one of the primary key performance indicators (KPI) for airlines, and the LF of an airline route is one of the most critical unknowns when opening a new line. In addition, understanding the associated variables and their importance in predicting the LF is essential for
making a managerial decision about the condition of a route. This study employed Naïve Bayes, Fast Large Margin, and Deep Learning models with Explainable AI (XAI) SHapley Additive exPlanations (SHAP) methods for local and global interpretations of how airlines could increase their LF. The Naïve Bayes model gave acceptable predictive performance results (98.1 % Mean ROC, 90.35% Precision, 77.08 Recall, and 91.5% Accuracy) and the most interpretable results. The preliminary results present relevant insights into understanding the factors that drive efficiency in the Brazilian aviation system.
making a managerial decision about the condition of a route. This study employed Naïve Bayes, Fast Large Margin, and Deep Learning models with Explainable AI (XAI) SHapley Additive exPlanations (SHAP) methods for local and global interpretations of how airlines could increase their LF. The Naïve Bayes model gave acceptable predictive performance results (98.1 % Mean ROC, 90.35% Precision, 77.08 Recall, and 91.5% Accuracy) and the most interpretable results. The preliminary results present relevant insights into understanding the factors that drive efficiency in the Brazilian aviation system.
| Original language | American English |
|---|---|
| Pages (from-to) | 39-62 |
| Number of pages | 24 |
| Journal | Journal of Supply Chain and Operations Management |
| Volume | 22 |
| Issue number | 1 |
| State | Published - 2025 |