Optimization Site Selection for Online-to-offline Convenience Facilities using Knowledge Graph and Graph Neural Network: A Case Study of Nanjing, China

Abstract

With the rapid growth of online retail and instant delivery services, traditional location models for urban facilities have undergone significant changes. This study proposes an optimization framework for the site selection of O2O (online-to-offline) convenience facilities in Nanjing, utilizing Knowledge Graphs (KGs) and Graph Neural Networks (GNNs). By integrating resident feedback, urban environmental data, and commercial facility performance, the framework identifies key environmental factors affecting facility operation and provides optimized spatial distribution recommendations for various O2O facilities. The findings show that the hybrid recommender system combining KGs and GNNs offers substantial advantages in capturing the complex relationships between the urban environment and facility performance, leading to more accurate and personalized site selection decisions. Compared to traditional methods, this approach enhances recommendation accuracy, interpretability, and the transparency of the site selection process. This study provides a scientific basis for improving the layout of O2O facilities, particularly in incorporating consumer feedback, thus enabling more human-centered decision-making in site selection.