<![CDATA[Improved system performance diagnosis of coronary heart disease becomes an important topic in research for several decades. One improvement would be done by features selection, so only the attributes that influence is used in the diagnosis system using data mining algorithms. Unfortunately, the most feature selection is done with the assumption has provided all the necessary attributes, regardless of the stage of obtaining the attribute, and cost required. This research proposes a hybrid model system for diagnosis of coronary heart disease. System diagnosis preceded the feature selection process, using tiered multivariate analysis. The analytical method used is logistic regression. The next stage, the classification by using multi-layer perceptron neural network. Based on test results, system performance proposed value for accuracy 86.3%, sensitivity 84.80%, specificity 88.20%, positive prediction value (PPV) 90.03%, negative prediction value (NPV) 81.80%, accuracy 86,30% and area under the curve (AUC) of 92.1%. The performance of a diagnosis using a combination attributes of risk factors,symptoms and exercise ECG. The conclusion that can be drawn is that the proposed diagnosis system capable of delivering performance in the very good category, with a number of attributes that are not a lot of checks and a relatively low cost.]]>
Copyrights © 2017
