Alternating decision trees for early diagnosis of heart disease

Abstract

Recent survey shows that heart disease is a leading cause of death in India and in world wide. Significant life savings can be achieved, if a timely and cost effective clinical decision system is developed. Adverse reactions occur if a disease is not diagnosed properly. A clinical decision support system can assist health care professionals for early diagnosis of heart disease from patient's medical data. Machine learning and modern data mining methods are useful for predicting and classifying heart disease. In this paper we wish to develop effective alternating decision tree approach for early diagnosis of heart disease. Alternating decision tree is a new type of classification rule. It is a generalization of decision trees, voted decision stumps and voted decision trees. We have applied our approach on heart disease patient records collected from various hospitals in Hyderabad. Optimization of features improves efficiency of earning algorithm. We used PCA to determine essential features of heart disease data. Experimental results show that our decision support system achieves high accuracy and proving its usefulness in the diagnosis of heart disease.