Research on Interpretable Recommendation Methods for Civil Aviation Ancillary Services

In the modern civil aviation industry, with the escalating market competition and the increasing diversification of airline profit models, revenue from ancillary services has become a key factor for airlines to enhance their overall revenue and profitability. Leveraging existing mature artificial intelligence technologies, insights into passengers' travel preferences and consumption patterns can be derived from their historical behavioral data, providing precise recommendations for ancillary services with interpretable output results to assist in decision-making processes. Based on this, a deep neural network with interpretability is proposed in this paper. By integrating full attention to capture hidden patterns in passenger behavior, the model's interpretability and recommendation accuracy are enhanced. Additionally, the use of residual structures effectively addresses the issue of gradient vanishing in traditional neural networks, subsequently improving recommendation accuracy and model convergence speed. Experimental validation on real datasets has demonstrated the effectiveness of the proposed approach. Furthermore, this paper provides an outlook on future research directions.