M.Sc. Dissertation - NIT Patna 2024

HealthGuard AI

Healthcare Insurance Fraud Detection

Ensemble of five Optuna-tuned classifiers (F1 = 0.7345) with SHAP/LIME explainability, secured by a SHA-256 blockchain audit trail with ECIES encryption (secp256k1 + AES-256-GCM).

Try Provider AnalysisView Research Results
0

Providers Analyzed

0

Best F1 Score

0

Blockchain Blocks

0

Engineered Features

System Pipeline

Ingest Claims

ECIES Encrypt

ML Ensemble

SHAP / LIME

Blockchain

Interactive Demonstrations

Provider Analysis

Look up any of 5,410 providers and view their fraud risk score with SHAP feature importance and LIME explanation rules

Open Demo →

Research Results

Complete experimental findings: model comparison, statistical significance (Friedman p=0.00089), bootstrap CIs, and 15 research figures

Open Demo →

Blockchain Explorer

Inspect the 110-block SHA-256 audit chain with Merkle tree verification, block details, and chain integrity validation

Open Demo →

ECIES Encryption

Live browser-side encryption demo using secp256k1 + HKDF-SHA256 + AES-256-GCM. Generate keys, encrypt, and decrypt data

Open Demo →

Technology Stack

PythonXGBoostLightGBMCatBoostSHAPLIMEOptunaSHA-256secp256k1AES-256-GCMNext.jsTypeScript

Dubasi Pavan Kumar - Supervised by Dr. Rajesh Kumar Sinha