KALP YETMEZLİĞİ HASTALIĞININ TAHMİN EDİLMESİNDE SINIFLANDIRICI ALGORİTMALARININ PERFORMANSLARININ DEĞERLENDİRİLMESİ

Abstract

Success rates and performances of Gaussian Naive Bayes, Support Vector Machines, Linear Discriminant Analysis, Decision Tree and Random Forest classifier algorithms from machine learning methods were evaluated using the Heart Failure Prediction dataset. Label encoder method was used primarily in data preprocessing techniques on the data set. Catalog data (5 pieces) in the data set have been converted into numerical data. In addition, it was observed that there were negative values in the data in a field and this situation was converted to values in the range of 0 - 1 with min-max conversion methods. After the pre-processing, analyzes were made with classification algorithms. As a result of these analyzes, a success rate of 90.76% was achieved with the random forest algorithm, which is an ensemble classifier. In the study, 80% of the data was used for training and 20% for testing. Of the 184 data used for the test, 102 of them were patients with heart failure and 72 of them were from those without the disease. The success of the random forest algorithm in estimating those with heart failure disease was 93.1% (95 observations), and the success in predicting those without the disease was 87.8% (72 observations).