Implementation and Optimization of a Fuzzy Rulebased Classifier for Multi-Class Classification Using Horizontal Federated Learning

In this paper, we investigate the practicality of a fuzzy rule-based classifier in horizontal federated learning and conduct optimizations. We perform simulations in order to evaluate the performance of different methods, such as accuracy, convergence rate and communication over-head etc. Our results show that the classifier effectively utilizes distributed data, which means it achieves better accuracy when more nodes are involved and conversely increases communication overheads and resource utilization. The model is robust to data distribution skewness and noise, with high-performance results should you use different class numbers as another example. These results illustrate the classifier trade-off between accuracy-in-human, interpretability and computational efficiency that makes it a well-rounded tool for federated learning scenarios. Further work on communication strategies and resource administration, as well as doubtlessly different city responses must follow which critically consider realworld application.