Medical Diagnosis of Human Heart Diseases with and without Hyperparameter tuning through Machine Learning

In medical field, cardiac disease prediction remains as a complex task. Heart disease is the major cause of death around the world. This study has surveyed various research papers and understood the methodologies used by various authors, their requirements and the challenges faced in the implementation stage. The exisiting research work on heart disease prediction uses machine learning models like Logistic Regression (LR), Random Forest (RF), Support Vector Machine (SVM), K-nearest neighbor (KNN) to get the accurate results. The present study focuses on hyperparameter tuning. Hyperparameters vary from each model and the best parameters will also change according to the dataset. By using the best parameters in learning model can give the prediction results in less time. Tuning of best hyperparameters can be done by using GridSearchCV. This study develops a web application by obtaining the input from the user. Features of heart diseases are retrieved from the historical data (the heart dataset). The major challenge is that the models mostly fails to give accurate results. Sometimes Hyperparameter tuning will give accurate result and in some cases it won't. It depends upon the type of model selected for model building. The proposed system mainly concentrates on comparision of models with and without Hyperparameter tuning based on the accuracy.