An Investigation of Hyperparameters Optimization and Feature Reduction Techniques: Predicting Airline Passenger Satisfaction Using Machine Learning Models

Aviation companies faced significant losses due to transportation restrictions amid the COVID-19 pandemic. Their inability to provide customer-friendly services resulted in a decrease in customer satisfaction. Machine learning models have become indispensable tools for predicting customer satisfaction and improving the quality of airline services. This study focused on utilizing three distinct feature reduction algorithms to generate three sets of feature collections, which were employed in six different machine learning models along with a hyperparameter tuning method, such as GridSearch, to achieve optimal results. GridSearch systematically evaluated numerous hyperparameter combinations. After applying three distinct algorithms for feature reduction and hyperparameter tuning, a comparative analysis of six different machine learning models was conducted to identify the one that provides the most accurate predictions of passenger satisfaction. The study’s outcome indicated that the Random Forest stood out as the most favorable choice for the highest performance score. It achieved an outstanding accuracy score of 97%, as well as impressive precision, recall, and F1-score values of 97%, 97%, and 97%, respectively. This model’s exceptional performance surpassed most recent similar studies regarding accuracy and efficiency.