Federated learning research: trends and bibliometric analysis

Federated learning (FL) allows machine learning algorithms to gain insights into a broad range of datasets located at different locations, enabling a privacy-preserving model development. Since its announcement in 2016, FL has gained interest from a variety of entities—both, in academia and industry. To understand what are the research trends in this area, a bibliometric analysis is conducted to objectively describe the research profile of the FL area. In this regard, 476 documents written in English were collected through a thorough systematic search in the Scopus database and examined from several perspectives (e.g., growth trends, top-cited papers, subject area), productivity measures of authors, institutions, and countries. Further, a co-word analysis through VOSviewer was carried out to identify the evolving research themes in FL. There has seen exponential growth in FL literature since 2018. There are five research themes, namely internet of things, wireless communication, privacy and security, data analytics, and learning and optimization, which were surfaced in the analysis. We also found that most of the documents related to FL were published in computer science, followed by engineering disciplines. It was also observed that China is at the forefront in terms of the frequency of documents in this area followed by the United States of America and Australia.