Passenger Satisfaction Prediction Model for Ethiopian Airline Utilizing Machine Learning Algorithms
No Thumbnail Available
Date
2025-08-01
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
Passenger satisfaction prediction is a critical task for Ethiopian Airlines to maintain service excellence and sustain a competitive advantage. Simple Traditional methods, such as statistical tools, often lack the analytical depth required to uncover complex patterns in large-scale passenger data. This study aims to develop a machine learning-based predictive model capable of assessing passenger satisfaction proactively, allowing Ethiopian Airlines to implement timely improvements and maintain service excellence. To conduct this study, passenger data was collected from Ethiopian Airlines' customer survey system, resulting in a dataset of 20,000 records between January 2022 and January 2023. Each response includes ratings on service quality, delays, staff interaction, and overall satisfaction, along with demographic details such as age and travel class. Data was gathered via both digital and paper surveys distributed at various touchpoints, including check-in, baggage claim, and post-flight followup. The dataset’s key attributes include passenger ratings, delay durations, demographic variables, and service-related feedback, providing comprehensive metadata for analyzing passenger satisfaction After acquiring the data, several preprocessing steps were applied to ensure data quality and compatibility with machine learning algorithms. Subsequently, we selected machine learning models that aligned with the objectives of the study and trained them to uncover the key factors influencing passenger satisfaction. The models were also used to evaluate the predictive performance of different algorithms. The dataset was split into 80% for training and 20% for testing. Finally, the trained models were evaluated to compare their accuracy and to identify the most influential features affecting passenger satisfaction. Based on the experiments conducted, ensemble models particularly XGBoost and Random Forest demonstrated the highest performance, achieving approximately 89% accuracy, meanwhile the deep learning algorithm (MLP, TabNet and DCN) achieves 85% accuracy in predicting passenger satisfaction. Ensemble models like XGBoost and Random Forest circumvented some of the common challenges faced by deep learning approaches such as data requirements, complexity, and interpretability thereby achieving higher predictive accuracy and providing more transparent insights into passenger satisfaction drivers. The dataset used in this study contained 20,000 records with 22 independent variables. Such data is sufficiently rich for ensemble models to learn complex non-linear relationships but not large enough to fully exploit the representational capacity of deep learning models (MLP, TabNet, and DCN) lacked sufficient data to optimize their numerous parameters, leading to slightly lower performance compared to ensemble methods. The most significant features found to be Type of Travel, Class, Customer Type, and Total Delay. This study confirms again that ensemble models like Random Forest and XGBoost are very effective for classification issues. Features such as Type of Travel, Class, Customer Type, and Total Delay were more influential on passenger satisfaction than traditionally assumed features such as Booking Service, Baggage Service, and Lounge Service. This result highlights the strength of machine learning and explainable AI (XAI) techniques in uncovering hidden patterns and identifying the real drivers of customer satisfaction in big and complex datasets
Description
Keywords
Passenger Satisfaction, Machine Learning, Deep Learning, Xg Boost, Random Forest, Explainable Ai