MVC in machine learning: a decade of algorithmic advances, challenges, and applications–a systematic review

Abstract

This systematic review evaluates the developments in multi-view clustering (MVC), its challenges, and applications from 2009 to 2024 and synthesizes 157 studies selected according to preferred reporting items for systematic reviews and meta-analyses (PRISMA) 2020 guidelines. MVC overcomes the shortcomings of the traditional single-view approaches by using complementary information provided by heterogeneous data sources. We used a strict search strategy in the ACM Digital Library, IEEE Xplore, and Scopus, and then carefully examined the quality of the found articles. The significant results suggest that the MVC research has grown explosively, with China as the major contributor and IEEE/Elsevier as the leading publishers. Developments in algorithms include deep learning, graph-based models, and factorization. Ongoing issues include managing incomplete views, scalability, successful fusion strategies, and interpretability. The review points out the wide range of applications of MVC in various areas, including bioinformatics, social network analysis, and multimedia. Future research must create adaptive frameworks, improve the interpretability of models, and develop strong evaluation measures, thus unlocking the full potential of MVC in real-life data applications.