Effective Heart Disease Prediction Model Through Voting Technique

Abstract

Machine learning has various practical applications that solves many issues relating to various domains .One among such domain is the health care domain and the most common application of machine learning is the prediction of an outcome based upon existing data in health care industry. Machine learning is shown as an effective technique in assisting the health care industry to make intelligent and effective decisions. The model tries to learn pattern from the existing dataset and later on it is applied to the unknown dataset for effectively predicting the outcome. Classification is the most effective technique for prediction of outcome. There are many classification algorithms which are used for prediction but only few algorithms predict with good accuracy whereas remaining algorithms predict with less accuracy. So to improve the accuracy of weak algorithms this paper presented a new method called ensemble classification ,where the accuracy is enhanced by combining multiple classifiers and later prediction is done by voting technique. So, experiments were done on a heart disease dataset, through ensemble approach the accuracy was enhanced and along with that a GUI was developed where the user himself can check whether he has probability of getting heart disease or not. The results of the study showed that ensemble method such as voting technique played a key role in improving the accuracy prediction of weak classifiers and also identified risk factors for occurrence of heart disease. An accuracy of 90% was achieved with voting technique and the performance of the process was further enhanced with a feature selection implementation, and the results showed significant improvement in prediction accuracy.

Two-level boosting classifiers ensemble based on feature selection for heart disease prediction

“…Ensemble learning methods like boosting and bagging (or bootstrap aggregation) are commonly utilized in classification tasks that involve manipulating training data [17]. Bagging is a common ensemble classifier technique in which several predictors are made separately and combined using model averaging methods, like the majority vote or the weighted average, to make a single prediction.…”