Efficient Classification of Heart Disease Forecasting by Using Hyperparameter Tuning

Abstract

Heart Disease is one of the serious diseases in the world. So, there is a huge requirement for early prediction and diagnosis of heart disease. Deep Learning is an emerging technology used widely in Health Care sector. In this research work, a sequential neural network model is used with corresponding parameters to develop heart disease prediction system. The performance of a model is assessing with different metrics like Accuracy, Precision, Recall and F1-Score. As values of parameters has effect on the performance of a model so choosing best optimal values is necessary for improvement. In this regarding, Hyper parameter tuning is used. For all parameters, every possible value will be tried to create a model and then their performance is evaluated. Manually, this evaluation is not that much effective and even become tedious and complex if number of parameters are high. So Gridsearchcv approach is used to tune parameters of model to get best set of values. Then these tuned values are used to create a model and then tested on test dataset. Finally model performance is evaluated with respect to corresponding evaluation metrics. From the evaluated results, it is observed that a model with Hyper parameter tuning induced highest accuracy of 83.60 than a model without Hyper parameter tuning.