Document Type

Article

Source of Publication

Procedia Computer Science

Publication Date

1-1-2024

Abstract

This paper addresses the critical challenge of fraud detection in medical insurance claims, a pervasive issue causing significant financial losses in healthcare. The primary goal is to develop an advanced fraud detection approach by integrating multiple unsupervised machines learning algorithms, leveraging their collective strengths through a majority voting mechanism, where labelling of data is unavailable. Central to this approach is the ensemble of 18 novel unsupervised algorithms, specifically, anomaly detection models. The novelty lies in the majority voting system employed to aggregate the decisions from these diverse algorithms, enhancing the reliability and accuracy of fraud detection. To validate the effectiveness of the proposed system, a dual approach is employed. Firstly, human experts in the medical insurance field review a subset of claims to establish a benchmark for the model's performance. Secondly, the system's effectiveness is quantitatively assessed using key statistical metrics. The system utilizes real-world insurance claim data to ensure quality and relevance, where the two datasets were collected from countries in the Gulf region. The findings reveal significant improvement in fraud detection at various activity levels; from doctor, provider, and patient, where the patient model reached 79 % precision. The system not only aligns well with human expert judgments but also demonstrates superior performance on the specified statistical metrics, indicating effective clustering and anomaly detection. Some real use cases were captured by the model and deeply investigated by human experts, which demonstrated advantages by the proposed approach in detecting fraud at multiple levels, of providers, doctors, and patients.

ISSN

1877-0509

Publisher

Elsevier

Volume

244

First Page

9

Last Page

22

Disciplines

Computer Sciences

Keywords

Fraud detection, Medical insurance, Unsupervised algorithms, Majority voting, Anomaly detection

Indexed in Scopus

no

Open Access

yes

Open Access Type

Gold: This publication is openly available in an open access journal/series

Download

Share

COinS