A Two Level Multi-Parametric Classification Approach for Prediction of Coronary Heart Disease

Abstract: Prediction system is one of the major research area under data mining. It is useful to perform the necessary action on time so that the future loss can be minimized. In this paper, we have defined a prediction system to identify the chances of occurrence of coronary heart disease based on person’s basic information, symptoms as well as medical tests. The medical disease prediction system is an application of expert system, that we here define by using an intelligent soft- computing approach called neuro-fuzzy. In this paper, we have defined a two layered approach for identifying the disease possibility. The critical factors that are mandatory for occurrence of coronary heart disease are taken at first level and the rest one are taken at second level. This two level approach increases the performance of our work as it helps in predicting disease chances accurately. The heart disease dataset is taken from UCI machine learning repository to train the neural network and then fuzzy rules are applied to predict the chances of coronary heart disease as low, medium or critical.

Keywords: Heart disease, Neuro-fuzzy, Coronary heart disease, Classification, Data mining, Back-propagation Algorithm