Exploring Federated Learning with Naïve Bayes using AVC Information
Mohammad Masudur Rahman, Dewan Md. Farid · 2023
Emerging trends in federated learning have drawn tremendous attentions in recent years where multiple clients collaborate for building a shared learning model without exchanging their own data. This approach ensures privacy, security and ownership of data, making it prevalent for large-scale data analysis. There have been a growing tendency to utilize deep neural networks in federated learning. However, many situations overwhelmed, and they require simpler learning algorithms along with prevention of data leakage. In this work, we propose a federated learning system based on Naïve Bayes algorithm utilizing AVC set. Naïve Bayes is a simple but computationally efficient probabilistic classifier. We leverage the strengths of Naïve Bayes to provide a secure and scalable solution for classification tasks in federated learning environments. To demonstrate the performance of our proposed approach, we conducted experiments on ten real-world benchmark data sets from the UC Irvine Machine Learning Repository. The results of the experiment showed that the Naïve Bayes algorithm enhances privacy preservation and scalability in a federated environment without compromising prediction accuracy compared to centralized machine learning. This study highlights the potential of using Naïve Bayes in federated learning and offers insights into its practical applications in various domains.