Machine Learning Techniques for Predicting Heart Diseases

Cardiovascular disease is another term for heart disease, refers to a variety of heart-related illnesses and has been the leading cause of death in recent decades. In the healthcare industry, data mining is a typical tool for evaluating enormous amounts of data. Wide variety of data mining and machine learning methods were used to analyze large amounts of complex medical data, assisting healthcare providers in identifying heart diseases. This works aims to used Sequential Minimal Optimization (SOM), Naive Bayesian, Random Forest, Stochastic Gradient Descent (SGD), Multilayer Perceptron, Adaboost, and Logistic Regression as classifiers and evaluate it accuracy with Heart Stalog Dataset. To improve performance of the selected classifiers Harmony Search Algorithm (HSA) was used as effective features selection tools. Feature selection is a key topic for dimension reduction and classification in high-dimensional datasets. During the feature selection procedure, only the most relevant attributes from the datasets will be chosen. The results show the accuracy was enhanced in Random Forest from 87.037% and AdaBoost from 84.0741% to 92.123% and 94.234 % respectively after using HSA on least number of attributes.