The project successfully developed a RAG-based chatbot prototype for P2P lending, integrating real-time borrower interactions with an XGBoost risk model. The chatbot effectively assessed loan applications, provided personalized feedback, and improved decision-making automation. Key achievements include:

Enhanced User Interaction: The chatbot dynamically adapted conversations, maintaining contextual relevance.

Automated Risk Assessment: Real-time borrower input improved loan evaluation accuracy.

Continuous Learning: Feedback loops refined the model, enhancing prediction quality.

Scalability & Feasibility Insights: Identified challenges in deployment, data privacy, and regulatory compliance.

The results provide a strong foundation for further optimization, with a focus on scalability, regulatory alignment, and expanded AI-driven financial services.

The project developed a chatbot prototype leveraging Retrieval-Augmented Generation (RAG) to enhance peer-to-peer lending decisions while minimizing common hallucination issues in large language models. The chatbot engaged users in interactive loan assessments, integrating real-time borrower input with a machine learning model based on XGBoost for risk evaluation.

Key achievements include dynamic conversational capabilities, automated decision-making enhancements, and feedback loops for continuous learning. The study provided valuable insights into user interaction patterns, operational scalability, and the feasibility of real-world deployment, laying the groundwork for further development and regulatory compliance integration.