Abstract
The rapid growth of multi-source, multi-perspective data in healthcare, finance, and social media has increased the need for unsupervised learning methods that integrate diverse views while preserving data privacy and locality. Federated learning (FL) enables collaborative model training without sharing raw data, yet its application to multi-view clustering (MVC) remains underdeveloped. This survey presents the first comprehensive, PRISMA-guided systematic review of federated multi-view clustering (FedMVC). A multidimensional taxonomy classifies existing studies according to FL paradigms (horizontal, vertical, and federated transfer learning); clustering approaches (hard/soft, spectral, density-based, deep, tensor, and non-negative matrix factorization [NMF]-based); aggregation strategies (FedAvg, FedProx, FedOpt, and adaptive optimizers); privacy mechanisms (differential privacy, secure multi-party computation, homomorphic encryption, and trusted execution environments); and non-IID data-handling techniques. Centralized and federated objective functions are examined, with comparisons of computational complexity, communication overhead, privacy guarantees, and scalability. Practical implementation guidance and applications in healthcare, finance, and cross-platform social media are also discussed. Key research challenges include formal privacy–utility trade-offs, convergence under extreme data heterogeneity, and the integration of large language models for improved interpretability. Overall, FedMVC represents a promising framework for privacy-preserving unsupervised learning in distributed environments.