A Machine Learning Based Expert System for The Diagnosis of Cardiovascular Diseases

Early detection of cardiovascular diseases (CVDs) can save hundreds of thousands of lives. Several AI-based CVD diagnosis methods have been developed without sufficient validation. These methods face several major challenges: 1) deleting instances from datasets with missing values, which may include significant information; 2) not utilizing any/proper feature selection technique to select relevant features, which can greatly enhance diagnosis task. To overcome these limitations, we present three intelligent, efficient methods to classify the risk of having CVDs. Two well-known missing value imputation methods were used, followed by three feature selection algorithms. Chi-square test (CHI), Least Absolute Shrinkage and Selection Operator (LASSO), and Genetic Algorithm (GA) have been applied as feature selector. For classification Multilayer Perceptron with backpropagation (MLP-BP) algorithm has been used. These models are then tested using four real world datasets. After that, the results are compared with other existing models.