IFLNLP: An Improved Federated Learning and Natural Language Processing-based Smart Healthcare System

The healthcare industry has automated due to artificial intelligence (AI). The smart healthcare system has been greatly improved by current developments and significant capabilities of machine learning (ML), deep learning (DL), and natural language processing (NLP). Still, there are certain challenges left that need to be addressed. The research of privacy-preserving, decentralized learning strategies is necessary for NLP and ML problems because of increasing rules and regulations. Federated learning (FL) provides straightforward methods for multiple clients to work together to train a single global model that meets the objectives of all clients while protecting local data, offering continuous training, and building a useful platform to handle heterogeneous input from various models. The proposed study presents a novel Improved-FLNLP (IFLNLP) and analyzes NLP workflow for smart healthcare with a special focus on FL. IFLNLP is an analytical framework for FL and NLP based smart healthcare systems. The experimental results obtained prove that the proposed approach, IFLNLP, yields the highest level of data privacy $\mathbf{(9 8 \%}$) and cost effectiveness $(\mathbf{9 5 \%})$, showing its strong privacy protection skills and better cost effectiveness as compared to NLP and FL. It maintains the same accuracy rate with the FL-based system and comparatively better accuracy than the NLP-based system.