GENERATED ARTIFICIAL INTELLIGENT SYSTEMS FOR AIRLINE TICKET MANAGEMENT

ABSTRACT The Airline Ticket AI system leverages artificial intelligence to enhance airline customer service and pricing by automating query handling and enabling dynamic price predictions, addressing inefficiencies in traditional systems such as slow manual responses and static pricing. This undergraduate project developed a conversational chatbot using a fine-tuned DistilBERT model to classify customer queries into 10 categories (e.g., price inquiries, cancellations) with 98% accuracy, incorporating context-aware responses through a rule-based dialogue manager. A Linear Regression model predicts ticket prices based on day, demand, and distance, achieving a Mean Absolute Error of ~$34.15 (17.07% of the average price, $200). The system is integrated into a user-friendly Streamlit web interface, hosted locally, with a load time of ~6 seconds and query processing in ~1 second on 8GB RAM, CPU-only hardware. Synthetic datasets (1,000 queries, 1,000 pricing records) were preprocessed for training, with iterative refinements resolving challenges like infinite loops, model file compatibility (model.safetensors), and PyTorch-Streamlit conflicts. The hybrid Waterfall-Agile methodology ensured structured development and adaptability. Results demonstrate effective automation and pricing flexibility, outperforming traditional systems in speed and context-awareness. The system offers a scalable, cost-effective solution for airline management, with recommendations for real-world data integration and cloud deployment to enhance future applicability.