A Knowledge Graph-based Retrieval-Augmented Generation Framework for Algorithm Selection in the Facility Layout Problem

Abstract: Selecting a solution algorithm for the Facility Layout Problem (FLP), an NP-hard optimization problem with a multiobjective trade-off, is a complex task that requires deep expert knowledge. The performance of a given algorithm depends on specific problem characteristics such as its scale, objectives, and constraints. This creates a need for a data-driven recommendation method to guide algorithm selection in automated design systems. This paper introduces a new recommendation method to make such expertise accessible, based on a Knowledge Graph-based Retrieval-Augmented Generation (KG RAG) framework. To address this, a domain-specific knowledge graph is constructed from published literature. The method then employs a multi-faceted retrieval mechanism to gather relevant evidence from this knowledge graph using three distinct approaches, which include a precise graph-based search, flexible vector-based search, and high-level cluster-based search. The retrieved evidence is utilized by a LLM to generate algorithm recommendations with data-driven reasoning. The proposed KG-RAG method is compared against a commercial LLM chatbot with access to the knowledge base as a table, across a series of diverse, real-world FLP test cases. Based on recommendation accuracy and reasoning capability, the proposed method performed significantly better than the commercial LLM chatbot.