Implementation of Artificial Neural Networks and Decision Tree Algorithms for Heart Disease Diagnosis

...

Abstract

An abnormal condition which troubles a living organism is called a disease. Nowadays, the most common problems the people are affected by are the heart problems. Several times, they lead to death in most cases due to the lack of correct diagnosis. The volume of data has been increasing rapidly in the area of health care. Predicting the heart problems is very difficult for the physicians. It is intractable to find the interesting patterns among enormous volumes of data. To find those, pattern recognition can be used, and to discover the hidden knowledge, data mining can be used. There have been a large number of medical data sets available in the market. Among all types of heart diseases, Cardio Vascular Disease is a type. So, many researchers carried out their works in heart disease dataset with 13 attributes, and 15 attributes with various data mining methods. In this study, ranking method was used in preprocessing a stage with total of 17 attributes for strengthening the rate of accuracy. The Zero R and J48 algorithms from NN and Multilayer Perceptron & decision tree were applied respectively on the dataset. The classifiers’ performance was analyzed by error rate and time complexity with accuracy. In this research, Multilayer perceptron classifier showed high accuracy results with 13 attributes. Out of these three classifiers, J48 classifier gave high accuracy, minimum error rate and less time while using 17 attributes. Hence, these approaches can be very useful to the physicians to take decisions at the proper time. This research work was entirely carried out by WEKA (Waikato Environment Knowledge Analysis) data mining tool.