PREDICTION OF HUMAN HEART DISEASE

Data mining techniques have been widelyused in clinical decision support systems for predictionand diagnosis of various diseases with good accuracy.These techniques have been very effective in designingclinical support systems because of their ability todiscover hidden patterns and relationships in medicaldata. One of the most important applications of suchsystems is in diagnosis of heart diseases because it is oneof the leading causes of deaths all over the world. Almostall systems that predict heart diseases use clinical datasethaving parameters and inputs from complex testsconducted in labs. None of the system predicts heartdiseases based on risk factors such as age, bloodpressure, fasting blood sugar, chest pain etc. Heartdisease patients have lot of these visible risk factors incommon which can be used very effectively for diagnosis.System based on such risk factors would not only helpmedical professionals but it would give patients awarning about the probable presence of heart diseaseeven before he visits a hospital or goes for costly medicalcheck-ups. Hence this paper presents a technique forprediction of heart disease using major risk factors. Thistechnique involves two most successful data mining tools,Support vector machine and Principal componentanalysis. The hybrid system implemented uses the globaloptimization advantage PCA for initialization of neuralnetwork weights. The learning is fast, more stable andaccurate as compared to back propagation.