View Proposal

---

Proposer

Heba Elshimy

Title

Explainable AI for Detecting Fraud in Health Insurance Claims

Goal

Build a machine learning model that flags potentially fraudulent healthcare providers and explains each flag in terms an auditor can act on.

Description

Healthcare fraud costs insurers billions each year. A flagged provider faces an audit, so every flag must be justified. In this project you will link US Medicare prescribing data with the government's list of excluded providers to build a labelled dataset. You will then train classifiers such as logistic regression, random forest and XGBoost to detect suspicious providers. Finally, you will apply explainable AI methods such as SHAP and counterfactual explanations to show why each provider was flagged. Key challenges are severe class imbalance and avoiding data leakage around exclusion dates.

Resources

Datasets: CMS Medicare Part D Prescribers by US Government, and the OIG LEIE exclusion list. The Kaggle Healthcare Provider Fraud Detection dataset is a fallback. Tools: Python, pandas, scikit-learn, XGBoost and SHAP. Google Colab or a standard laptop is enough; no GPU is needed.

Background

Python programming and core machine learning (classification, evaluation metrics), for example F21DL. No healthcare knowledge is required.

Url

External Link

Difficulty Level

Moderate

Ethical Approval

None

Number Of Students

1

Supervisor

Heba Elshimy

Keywords

machine learning, fraud detection, ai in healthcare

Degrees

Bachelor of Science in Computer Science
Bachelor of Science in Computer Systems
BSc Data Sciences