Parameter recommendation system enhanced by RAG based on process knowledge

This article proposes an innovative method that utilizes the Neo4j graph database as a vector database to segment text, images, and table contents in process files and convert them into vector form for storage. On this basis, a parameter recommendation system was designed and implemented. The system efficiently retrieves parameter reference questions raised by users through vector similarity matching technology, and passes the matched relevant results to advanced machine learning models for further analysis. Ultimately, the system is able to accurately answer users' questions and recommend corresponding parameter ranges, significantly improving the efficiency and accuracy of enterprises in obtaining specific parameter requirements.