A Deep Learning-Based Classification Model for the Detection of Heart Diseases

ABSTRACT

This study focused on the development of a classification model needed for the detection of heart diseases based on information about relevant features using a deep neural network (DNN) architecture. Data was collected from a public data repository provided by Kaggle following which the data was preprocessed and feature selection was used to select the most relevant features. A DNN architecture was used to implement the classification model based on information about initially identified features and relevant features. The classification models generated from these datasets were compared using evaluation metrics. The results of the study revealed the level of importance of the features associated with the detection of heart diseases using the value of mutual information and an improvement in the performance of models based on such features. The study concluded that information about relevant features provide better performance than dataset that may contain information about irrelevant data.

Keywords: Classification, Deep learning, Heart diseases, Predictive modeling, Feature Selection.