A Federated Recommendation Method for Personalized Manufacturing Service Composition Considering Collaborative Relationship

In the industrial Internet environment, the complexity of manufacturing tasks is gradually increasing and can no longer be completed by independent manufacturing services. Most recommendation systems have drawbacks in data security when exchanging client parameters and do not adequately consider individual preference differences. Therefore, this paper proposes a federated recommendation method for personalized manufacturing service composition considering collaborative relationships (CoFedSVD++) among enterprises. First, the cosine similarity method is improved to measure the QoS correlation between manufacturing services. Second, a homomorphic encryption algorithm based on the federated learning framework is established to hide the model parameters. Third, the improved SVD++ algorithm discovers individual personalized preferences through implicit feedback information. Finally, a real industrial Internet platform case is adopted to compare the recommendation effect of different recommendation algorithms.