Recommendation System addresses an analytical framework for automatically collecting and refining diverse product data, and intelligently recommending competing products based on it. This study aims to overcome the inefficiency and expertise variability of traditional manual market research by integrating a data pipeline with an LLM-based analytical architecture. In particular, it targets explainable and reliable competitor recommendation grounded in real data through a RAG-based recommendation system.

• Problem Inefficiency of manual product analysis and expertise variability in competitor discovery.
• Pipeline Automated pipeline for collecting and refining product metadata and unstructured data.
• Approach LLM-based category mapping and data consistency validation.
• System Hybrid retrieval-based RAG system combining dense, sparse, and re-ranking methods.
• Goal Explainable and reliable evidence-based competitor recommendation system.