Federated Learning for Privacy-Preserving Healthcare Data Analysis
Students & Supervisors
Student Authors
Supervisors
Abstract
The present research aims to study, we focus on the utilization of federate learning techniques for analyzing the health care data of Bangladeshi citizens. It particularly aims at assessing the viability of FL as a method to build accurate and robust machine learning models under privacy concerns. Using a time-series dataset over some time (2016 – 2023), significant measures were analyzed, including the degree of medical services establishments that embrace FL, the quantity of pilot projects, the spread of patient data, the reduction of the episode of information breaks, and the precision of the model. The data came from the ICT Division, the Directorate General of Health Services, academic publications, and reports from non-governmental groups. Their use grew, from 0.5%, in 2016 to 20% in 2023, and the number of pilot projects expanded from 1 to 20. 50% More Patient Data; +50% Less Data Breaches The accuracy of machine learning models increased from 72 per cent in 2016 to 87 per cent in 2023 helped by improved algorithms and collaboration between institutions. This progress has been made possible by initiatives from governmental and non-governmental organizations. Federated learning enables Bangladesh to bring forth privacy-centric healthcare after tackling infrastructure issues and prohibitive upfront costs. This distributed method is, however, scalable and does enable cross-institution studies without leaking patient information in the face of hurdles like capital costs and remote connectedness.
Keywords
Publication Details
- Type of Publication: Conference
- Conference Name: 3rd National Mathematics Conference 2024
- Date of Conference: 06/02/2025 - 07/02/2025
- Venue: Bangladesh University of Engineering and Technology (BUET) Dhaka, Bangladesh.
- Organizer: Bangladesh University of Engineering and Technology (BUET) Dhaka, Bangladesh.