Higher-Order Network Structure Embedding in Supply Chain Partner Link Prediction

Enterprise partner link prediction is a research direction of the recommendation system, which is used to predict the possibility of links between nodes in the enterprise network, and recommend potential high-quality partners for enterprises. This paper is based on the automobile enterprise network, and study how to recommend high-quality parts suppliers for auto manufacturers, then propose a supply chain corporate partner link prediction algorithm embedding in higher-order network structure. The coupled rating matrix and triad tensor model is constructed by mining the higher-order link patterns in the enterprise network, and considering the interaction between user demand and automobile manufacturers, which explicitly reflects the auto manufacturer’s choice of its part suppliers. The model uses the Alternating Direction Multiplier Method (ADMM) to solve the problem, which effectively alleviates the data sparsity problem in the recommendation system. On real data crawled from automobile-related websites, experiments show that the algorithm can obtain more accurate link prediction effects than traditional algorithms.