Decentralised Federated Learning Framework for Privacy-Preserving Medical Insurance and Sustainable Rural Economic Development in Tamil Nadu
Abstract
Abstract Rural medical-insurance entrepreneurship in India is constrained by fragmented, siloed health and claims data that insurance agents, primary health centres, and cooperative micro-insurers cannot legally or practically pool. Federated learning (FL) is proposed as ecosystem-level infrastructure that could let such dispersed actors jointly train a shared underwriting-risk model without centralising raw data. This paper develops and rigorously evaluates TW-DP-FedAvg, a federated architecture combining gradient-clipped Gaussian differential privacy with trust-weighted client aggregation, on two real public benchmark datasets (Medical Cost Personal Dataset, n = 1,338; Pima Indians Diabetes Dataset, n = 768) partitioned into five Dirichlet-heterogeneous simulated district shards, since no public dataset of real rural Tamil Nadu insurance records exists. Responding directly to reviewer concerns about an earlier version of this work, three methodological corrections are made: (i) the differential-privacy budget is now calibrated via the advanced composition theorem to a stated total epsilon of 8 across all 40 communication rounds, rather than reporting an uncomposed per-round figure; (ii) a 2×2 ablation isolates the independent contributions of the DP mechanism and the trust-weighting rule; and (iii) equivalence between the federated and centralised models is assessed with two one-sided tests (TOST) and Cohen's d, not only a non-significant p-value. Under the revised…
Read the paper
Related studies
- Active Learning during Federated Learning for Object Detection
- Fairness-aware federated learning
- Decentralized Federated Learning
- Building a Personally Identifiable Information Recognizer in a Privacy Preserved Manner Using Automated Annotation and Federated Learning
- Preservation of Patient Level Privacy: Federated Classification and Calibration Models
- Efficient explainable federated learning
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.