Enhancing Heart Disease Prediction Models with Feature Selection and Ensemble Methods

Pulugu Dileep, Kunjam Nageswara Rao and Prajna Bodapati

Heart disease prediction models based on data of patients have showed significant utility in early prediction of disease. Artificial Intelligence (AI) with machine learning algorithms yield required knowhow to determine heart diseases. Supervised machine learning has been around for prediction of diseases. Data used for prediction model may have irrelevant and redundant features. Feature selection and feature optimization models solve this problem by eliminating such features. This will improve prediction performance. Another optimization problem is the usage of ensemble of multiple classification models. In this paper we proposed a framework that considers ensemble of different prediction models in instruct to have enhanced prediction performance. In addition to this a feature selection algorithm named Heuristic Based Feature Selection (HBFS). Real world dataset is collected from Kaggle datasets resource. An experimental setup is made with Python environment with data mining package sk learn, keras and tensor flow. Anaconda is the data science platform used for empirical study. Different prediction models made up of Linear Regression, KNN, SVM, RF, DT, NB, NN and ensemble model. The empirical study revealed that the ensemble method and feature selection cloud provide enhanced prediction of heart disease.

Volume 11 | 11-Special Issue

Pages: 400-411

DOI: 10.5373/JARDCS/V11SP11/20193048