Airfare Fluctuation Analysis with Event and Sentiment Features by Stacking Ensemble Model

Fluctuations in airplane ticket prices are influenced by various factors, including market demand, seasonality, and major events. Events are particularly challenging to manage because their information cannot be directly obtained from flight data but must be sourced from external channels such as social media. To address this issue, this research integrates sentiment data from social media, specifically Twitter (X), with three major events: Lebaran, Vesak, and Eid al-Adha. These events were selected due to their significant cultural and religious importance, which has a considerable impact on travel demand and, consequently, ticket prices. The study utilizes a stacking ensemble model for sentiment analysis, combining Random Forest (RF), Support Vector Machine (SVM), and Voting Classifier (VC). This combination was chosen because RF effectively captures complex patterns, SVM excels in finding decision boundaries, and VC aggregates predictions for better stability and accuracy. The stacking ensemble model achieved an accuracy of $\mathbf{0 . 9 6 2 9}$, with a precision of 0.9273, recall of 0.9629, and an F1-Score of 0.9448, demonstrating the robustness of the approach. The sentiment results for each event were then merged with the airfare dataset, creating an explainable dataset with integrated event and sentiment features. These combined features, which are directly related to the events, were analyzed to identify and understand ticket price fluctuations. The analysis revealed significant differences in ticket prices before and after the events. By incorporating these features, the analysis of ticket price fluctuations becomes more accurate and informative, offering new insights into the dynamics of airplane ticket prices and aiding in strategic decision-making in the airline industry.