A Comparative Analysis of Federated Learning Towards Big data IoT with Future Perspectives

Over the past few years, big data has come a long way, allowing us to make sense of the immense amounts of information being produced by cutting-edge services and apps, as well as a plethora of IoT devices. Only through extracting value from huge data through data analytics can big data's full potential be accomplished [6]. Machine learning (ML) plays a highly essential part in this process because of its capacity to know patterns and then deliver insights through learning from facts. Since the data may include private information about individuals, governments, or financial accounts, using a more traditional technique raises serious concerns about data confidentiality. Recently, Federated learning (FL) seemed like a potential learning procedure that could help us get beyond this obstacle. In FL, the central ML algorithm is only given the parameters from the local devices so that it can train and make predictions on a global scale. The aforementioned problems with huge data may be amenable to resolution thanks to this aspect of FL. Furthermore, a thorough assessment of FL for big data applications has forward done, leaving a void in the literature. This article delivers a summary of FL, big data, and the motives for the procedure of FL for big data by surveying its use in big data applications. In particular, it is important to thoroughly examine the usage of FL for major big data services with applications like the smart city, remote surgery, transportation, smart grid, and social networks. Additionally, briefly discuss the most pressing problems associated with this engaging subject and the most potential approaches to fixing them.