Year: 2025
Type: conference-paper
Abstract: Due to the complexity of the current dairy supply chain and isolated data, traditional traceability methods are inefficient and cross-link transmission paths are difficult to identify when Salmonella breaks out. The purpose of this paper is to build a graph neural network model based on GraphSAGE to accurately locate the source of contamination and the transmission path. First, a heterogeneous supply chain graph is constructed: this paper takes dairy farms, processing plants, logistics vehicles, supermarkets, etc., as nodes, and the relationship between goods flow and personnel contact as edges; secondly, feature engineering is performed: nodes are embedded with Salmonella gene sequence data (SNP clustering), environmental temperature and humidity monitoring data, and enterprise quality inspection records. This paper trains GraphSAGE: a two-layer aggregation function (mean aggregator) is used, the number of sampled neighbors k=15, and the node representation is optimized after 20 rounds of iterations. It calculates the probability of cross-link transmission through node embedding and outputs high infection risk paths. In three real epidemic cases, the model traceability accuracy rate reached 94.7%, the recall rate was 92.1%. GraphSAGE significantly improves the efficiency of pathogen transmission path tracking in complex supply chains by integrating multi-source heterogeneous data and graph structure learning, providing decision support for rapid epidemic prevention.
Authors: Jing Sun, Xiangzhen Cui
Institution: Shandong Institute of Commerce & Technology
Language: English
Subfield: Food Science
Domain: Life Sciences
SDG: Zero hunger
Open Access status: gold
OpenAlex ID: W4414396241