Study of Various Classification Approaches including Deep Learning in Heart Disease Prediction

Abstract

During the last couple of decades, apart from other diseases, heart disease has proved to be a major cause of human death. This refers to a wide variety of heart conditions which include diseased vessels, structural problems and blood clots. This paper is aimed to make an effective prediction about the vulnerability to heart disease of a person depending on the given health parameters at a very early stage & to reduce premature death, with an improved accuracy and reliability compared to other traditional models. Our proposed prediction model has proved to be reliable in this connection and has yielded a maximum accuracy of 99.0% using Random Forest, Support vector machine and Decision trees and K-nearest neighbour algorithms. Using Cross Validation, we have also prepared the model (with the highest accuracy level of 98.6%) to work efficiently, taking the correct pattern of the dataset. Further, the proposed model has outperformed other traditional models in terms of accuracy using some other algorithms (Hybrid Ensemble model, Extreme Gradient Boosting with Random Forest and Stochastic Gradient Descent with accuracies of 94.2%, 87%, 78% respectively). Moreover, TensorFlow has also been used in order to get reliable prediction of heart disease with an approximate accuracy of 83.44%. The novelty of our proposed model lies in showcasing better results compared to those obtained by the traditional models and this model can easily be applied in medical science to provide better diagnosis.