Machine Learning Classification of the Coronary Artery Disease and Clustering of Free-Form Medical Complaints

Marinov, Boris (2021) Machine Learning Classification of the Coronary Artery Disease and Clustering of Free-Form Medical Complaints. Master's Thesis / Essay, Human-Machine Communication.

Preview	Text Master_Thesis_Boris_Final.pdf Download (3MB) | Preview
	Text toestemming.pdf Restricted to Registered users only Download (121kB)

Abstract

Cardiovascular diseases are some of the leading causes of death across the world, with the coronary artery disease (CAD) accounting for the highest mortality rates. Effective treatment relies on an early and accurate detection. The signs and factors of the disease are well understood, however often require expensive and invasive methods which are not easily accessible to most practitioners. Machine learning (ML) application to the field carries promising solutions to the problems, and is already a widely researched topic. The CONCRETE nationwide project was set up with the hopes of investigating whether ML-based prediction and analysis can be applied to easily obtainable, self-reported patient data. The data, despite being scarce, is made up of categorical quality-of-life answers and free-form, unstructured text complaints. Utilizing supervised method, this project shows that multi-label classification is possible to a degree when using the introspective answers only. Feature selection is used to quantify and discuss the gender and age-specific differences which contribute to the risk of the disease, promoting a more tailored future detection. Finally, unsupervised and NLP methods are combined to propose a form of topic modeling which goes beyond previous modeling methods and successfully discovers clusters of similar complaints for each disease group. The presented results pave a promising path and outline the future potential of the CONCRETE project.

Item Type:	Thesis (Master's Thesis / Essay)
Supervisor name:	Cnossen, F.
Degree programme:	Human-Machine Communication
Thesis type:	Master's Thesis / Essay
Language:	English
Date Deposited:	27 Oct 2021 12:01
Last Modified:	27 Oct 2021 12:01
URI:	https://fse.studenttheses.ub.rug.nl/id/eprint/26183

Actions (login required)

View Item