Effective Diagnosis of Coronary Artery Disease using Case-based Reasoning

Main Article Content

Yong-Gyu Jung, Bumsu Kim, Hojin Nam, Minseo Rhee, Jeung-Sun Lee


 With the advent of big data, data mining is more increasingly utilized in various decision-making fields by extracting hidden and meaningful information from large amounts of data. Even as exponential increase of the request of unrevealing the hidden meaning behind data, it becomes more and more important to decide to select which data mining algorithm and how to use it. There are several mainly used data mining algorithms in biology and clinics highlighted; Logistic regression, Neural networks, Support vector machine, and variety of statistical techniques. Among them Case-based reasoning (CBR) is relatively seems to be simplistic but very powerful to disclose unseeable problems in complex environments with only simplistic use of the above single technique for prediction of nonlinear models. On the other hand, quantities of the human momentum and activities are more diminished, whereas lifestyle of drinking, smoking and western eating habits are changing, and thus such as the unrevealed risks caused by heart attack or angina are growing up more and more. Therefore according to the increase of patients suffering from heart disease, a number of data mining studies are undergoing to assist medical doctors by prediction of whether to perform coronary angiography which requiring much resources in cost and procedures. Our study uses the same datasets on heart disease patients, that made use of multiple datasets collected from Cleveland, Hungary, Long Beach and Switzerland. Unlike the approach of , we observed that the experimental dataset is composed of multiple populations. And they are similar in use of same kinds of disease patients but different in the time and area of investigation. Through the experimental results, CBR made better performance than the techniques proposed from the original study for the disease prediction. Consequently we conclude effective diagnosis prediction must accompany with selection of the data mining technique considering the characteristics of samples and data collection.

Article Details