Cardiac arrhythmia classification using self organizing MAP (SOM) - based ensemble model

Dayang Yasmin, binti Abang Abdul Wahab (2015) Cardiac arrhythmia classification using self organizing MAP (SOM) - based ensemble model. [Final Year Project Report] (Unpublished)

Preview	PDF Cardiac arrhythmia classification using self organizing MAP (SOM)-based ensemble model (24 pages).pdf Download (1MB) | Preview
	PDF (Please get the password by email to repository@unimas.my , or call ext: 3914 / 3942 / 3933) Cardiac arrhythmia classification using self organizing MAP (SOM)-based ensemble model (fulltext).pdf Restricted to Registered users only Download (1MB)

Abstract

Many clinical decision support systems have been using data mining techniques for prediction and diagnosis of various diseases with good accuracy. This is due to its ability to distinguish various patterns of data from its background, and make conclusions about the categories of the patterns. A large number of such systems have been widely used in the diagnosis of heart diseases. One of the heart diseases in concern is cardiac arrhythmia. Most systems used in diagnosing cardiac arrhythmia uses data mining techniques, like Artificial Neural Networks, particularly in the form of a single classifier. In this project, a Self Organizing Map (SOM) - Based Ensemble model is proposed for the classification of cardiac arrhythmia disease dataset. An ensemble is a model that applies multiple learning models and combining the outputs or predictions to solve a particular problem. An ensemble is stated to predict or classify datasets more accurately than some single classifier models. The ensemble consists of three SOM classifiers trained with different number of dimension. For the ensemble, a voting technique is used to average the prediction of each single SOM classifier to obtain the final prediction. The results displayed show that the SOM ensemble model has higher classification accuracy than that of single SOM classifiers. Ensemble learning eliminates errors of single classifiers by averaging the prediction of each classifier, thus resulting in a more accurate output.

Actions (For repository members only: login required)

View Item