Imam Safii Heart Disease Classification using Gain Ratio Feature Selection with Hidden Layer Modification in Extreme Learning Machine

Heart disease is a non-communicable disease that causes a high mortality rate and is still a problem both in developed and developing countries. This disease often occurs because of the narrowing of blood vessels which causes the functioning of the heart is disturbed. The number of cases of heart disease in Indonesia is still quite high, making medical staff require a fairly in diagnosing the patient's conditional. The research proposed to implement Gain Ratio in selecting the most important feature that influences heart disease and building the classification models based on the modification of hidden layer weight on Extreme Learning Machine. The research collected the heart disease dataset which was obtained from Kaggle UCI Machine Learning consist of 1.025 samples, 14 attributes, and 2 labels. The data preprocessing include using data cleaning and normalization to find out dirty data or missing values. The experiment reported that Gain Ratio succeeds to generate the attribute ranking of heart disease dataset, then Gain Ratio score was added to the weighting of the hidden layer input on learning methods. The research used various validation sampling using the splitting test between training data and testing such as 70:30, 80:20, 90:10%, and set up 1500 hidden layers. The accuracy average performance of Extreme Learning Machine with modification using Gain Ratio reached 100% for the training phase and 97.67% for the testing phase.   Keyword: Heart Disease, Gain Ratio, Modification, Classification, Extreme Learning Machine