RAG를 이용한 마케팅 활용에 대한 연구 : 신용카드 할부 예측을 중심으로

초록

This study explores the potential applications of RAG based LLM in predicting installment purchases, a major revenue source for credit card companies. The research compared and analyzed the performance of various predictive models using both large and small datasets, with particular emphasis on validating the effectiveness of RAG-based LLM. Analysis results showed that the Random Forest model achieved the highest prediction accuracy in large-scale samples. Conversely, in small-scale samples, RAG-based LLM demonstrated superior predictive performance, particularly proving its high potential in analyzing new customer characteristics and developing personalized marketing strategies based on these insights. This study suggests that RAG-based LLM can serve as an effective tool that maintains comparable performance to existing predictive models while providing additional insights for marketing strategy development. Furthermore, the study derived practical implications by presenting the possibility of automating the "RAG - Analysis - Strategy Development" process using AI Agents.

키워드