Eco-Friendly AI-Based Classification System for Payment Card Support Tickets
- Proceedings of the 7th Artificial Intelligence and Cloud Computing Conference : 386-391
Résumé
The payment card industry has revolutionized the way we use and interact with money. The Card Payment Support Department is responsible for customer management, ensuring that bank cards, ATMs, POS terminals, and online stores function properly, allowing users to make transactions with their bank cards anytime and anywhere. The Payment Service Center serves as the first point of contact for customers, and its main challenges include the ability to quickly categorize customer requests, consistently resolve tickets within the SLA (Service Level Agreement), and enhance the user experience. In this paper, we propose a fine-tuned Large Language Model (LLM) to automatically classify incoming customer requests. Our model achieved an average accuracy of 76.74% across 51 categories, using only 3 epochs. The complete code and model are available online. We also developed a web application with Streamlit that leverages our model to predict the category of incoming tickets and automatically assign the request to the correct department when the model’s confidence is higher than 40%. Through this study, we demonstrate that it is possible to achieve good results in AI model training with minimal computational resources, which is a key consideration for the financial sector in West Africa and also beneficial for the environment.
Mots-clés
Payment, Smart card, Computer science, MULTOS, Payment card, Payment system, Computer security, Credit card, World Wide Web, Card security code