Investigations on cardiovascular diseases and predicting using machine learning algorithms

Abstract

Detection of heart diseases (HD) at an early stage diminishes the mortality rate. However, handling huge data is cumbersome for physicians. Hence, there is a need for a tool regarding the automation and processing of large data to help in making precise decisions. The proposed methodology aims to predict cases of HD or cardiovascular diseases (CD) well in advance based on the features of the patient. In the proposed methodology, the Cleveland Dataset of HD collected from University of California, Irvine (UCI) repository is evaluated. The various phases in the proposed methodology include, insight about the data, analyzing the data, feature engineering and finally building the model. The K-nearest neighbor (KNN), support vector machine (SVM), artificial neural network (ANN), classifier and convolutional neural network (CNN) are used to predict the HD and are evaluated on the dataset. The metrics considered are classification accuracy (CA), recall (r) and precision (p). The resulted CA, ‘r’ and ‘p’ for KNN-based approach are 66.7%, 91.7% and 88.8%, respectively. The SVM-based approach with ‘linear’ kernel achieved CA, ‘r’ and ‘p’ are 74.2%, 85.0% and 74.0%, respectively. The ANN-based approach resulted in 70.08%, 77.0% and 84.2% of CA, ‘p’ and ‘r’, respectively. Finally, the CNN-based prediction achieved CA, ‘p’ and ‘r’ are 83.61%, 76.0% and 97.0%, respectively. The experimental study concludes that the CNN-based prediction model outperformed the KNN, SVM and ANN-based prediction approaches regarding the prediction accuracy. An optimization algorithm can be incorporated into the model in future.

Keywords

• Heart disease
• prediction
• recall
• precision
• classification
• UCI repository