The role of machine learning in health insurance industry : a case study

Dissertant

Rashmawi, Ruba

Thesis advisor

Abu Hasan, Hasan

University

Birzeit University

Faculty

Faculty of Engineering and Technology

Department

Department of Computer Science

University Country

Palestine (West Bank)

Degree

Master

Degree Date

2022

English Abstract

Insurance companies worldwide are exploring how machine learning (ML) can improve customer satisfaction, reduce operational costs and increase profitability.

The greatest opportunity lies in risk prediction and fraud detection, whereas these two topics are the most challenging in the Insurance domain.

The first goal of the study is to predict the risk level in Health Insurance subscribers using machine learning Classification algorithms.

The study included around 10 thousand subscribers who were classified on 3 risk levels (High-risk, Mid-risk, and Low-risk).

The main objective is to predict accurately the risk level and accordingly assist the company in providing an accurate premium rate for new customers.

Some supervised classification algorithms were analyzed and compared using R statistical programming Language; the Random Forest algorithm had the highest accuracy which was around 94%, sensitivity for the High-risk level was around 72%, and 81% for the Mid-risk level.

The second goal of the study is to detect outliers in medical claims to help predict which claims are suspects of fraudulent activities and assist insurance specialists to investigate and discover cases of fraud.

Unsupervised Outlier Detection algorithms assist in discovering abnormal behavior in medical pattern claims and discover cases of fraud.

Three outlier detection techniques were compared, the first using Clustering algorithm (PAM), the second using Densitybased local outliers (LOF), and the third using Isolation Forests (IF).

Outliers with abnormal behavior were detected, some had very high costs compared to the clusters/neighbors they belong to, and others had a very small-time interval between clinical visits which may be suspicious behavior.

Claims audits consume lots of time and are very costly for insurance companies, so having machine learning algorithms detect the suspicious claims that require review will save the company huge amounts of operational costs and will increase work efficiency.

Main Subjects

Information Technology and Computer Science

No. of Pages

93

Table of Contents

• APA
• MLA
• AMA

Language

English

Data Type

Arab Theses

Record ID

BIM-1429254