Detection of heart failure with machine learning algorithms

Nowadays, where data has become important and meaningful, it has become inevitable for data in the field of health to transform into information in the face of developing technology. Cardiovascular disease is a disease that occurs due to many reasons such as the body's inability to pump the amount of blood it needs, structural and functional disorders of the heart, oxygen deficiency, and vascular occlusion. It is known that approximately 18 million people throughout the world die from cardiovascular diseases annually. Artificial intelligence algorithms have begun to be used in the field of medicine, as in many other fields, in order to assist hospital staff in the decision-making process and to minimize possible human errors. In this thesis study, a data set has been obtained by collecting 202 data from the Cardiology Department of Nevşehir State Hospital with the scale created by the ethical committee-approved Framingham criteria and specialist opinion. In addition to this data set, a data set containing 299 data created for heart failure called "Heart Failure Prediction", which is open access by the University of California, was also included in the study. In the study, five-way cross-validation has been applied to models created with Artificial Neural Networks,Decision Tree, Random Forest, Logistic Regression, K Nearest Neighbor, Naive Bayes machine learning algorithms on two data sets with a total of 501 samples. After the pre-processing process of the collected data, the performance results of the models created with the complexity matrix and accuracy, sensitivity, f1 score and precision criteria have been calculated. The accuracy values of the created data set have been achieved in Decision Tree (95%), Random Forest (93%), Logistic Regression (90%), Artificial Neural Networks (90%), K Nearest Neighbor (81%) and Naive Bayes (80%). Besides, with the WEKA program, the most important attributes affecting heart disease have been determined as Ejection Fraction, creatinine, anemia, shortness of breath, diabetes and Atrial Fibrillation.