Archives and Documentation Center
Digital Archives

A fuzzy support vector machine approach for ECG analysis

Show simple item record

dc.contributor Graduate Program in Systems and Control Engineering.
dc.contributor.advisor Gürgen, Fikret.
dc.contributor.author Özcan, N. Özlem.
dc.date.accessioned 2023-03-16T11:34:48Z
dc.date.available 2023-03-16T11:34:48Z
dc.date.issued 2010.
dc.identifier.other SCO 2010 O38
dc.identifier.uri http://digitalarchive.boun.edu.tr/handle/123456789/15655
dc.description.abstract In this thesis, fuzzy approach is used for ECG analysis. The ECG dataset in the UCI database is used. This dataset consists of inputs from normal and abnormal ECGs. All the anomalies are used to construct one class and normal ECG data is used to construct another class. The main purpose of the system is to detect anomalies correctly. In order to achieve this goal, a fuzzy support vector machine is constructed. Five different fuzzy membership functions are tested to reach the best performance: OCW, DTCM, DTOCM, CAR and FCM. Output of the fuzzy support vector machine system is compared to other classification methods. Results show that the fuzzy support vector machine outperforms other methods. In order to interpret the classification model, rule base extraction methods are applied. C4.5, PART, RIPPER and ANFIS are the selected algorithms for ECG rule base generation. When accuracy is considered as performance metric, RIPPER method outperforms the other techniques. The advantage of using ANFIS is the membership function generation for the features in the dataset. The resulting membership functions are found to be consistent with medical knowledge.
dc.format.extent 30cm.
dc.publisher Thesis (M.S.)-Bogazici University. Institute for Graduate Studies in Science and Engineering, 2010.
dc.relation Includes appendices.
dc.relation Includes appendices.
dc.subject.lcsh Electrocardiography.
dc.subject.lcsh Support vector machines.
dc.title A fuzzy support vector machine approach for ECG analysis
dc.format.pages xi, 119 leaves;


Files in this item

This item appears in the following Collection(s)

Show simple item record

Search Digital Archive


Browse

My Account