Hastalık tahmininde makine öğrenmesi sınıflandırma algoritmalarının karşılaştırılması ve bootstrap metodu kullanımı

GAMZE KABA

Hastalık tahmininde makine öğrenmesi sınıflandırma algoritmalarının karşılaştırılması ve bootstrap metodu kullanımı

Thesis Type: Postgraduate

Institution Of The Thesis: İstanbul Ticaret University, Fen Bilimleri Enstitüsü, İSTATİSTİK ANABİLİM DALI, Turkey

Approval Date: 2022

Thesis Language: Turkish

Student: GAMZE KABA

Supervisor: BAĞDATLI KALKAN SEDA

Open Archive Collection: AVESIS Open Access Collection

Abstract:

In the field of health, large data piles are formed with the recording of data for many years. Data stacks can be classified and made more understandable with using machine learning methods. These methods also allow the estimation of many disease diagnoses. In this study, various of risk factors for early diagnosis of Cardiovascular Disease, which is currently leading couse of death globally, were evaluated. Early diagnosis of the disease carries great importance in the field of health because it accelerates the treatment process. The dataset used in this study, consists of data accumulate under 11 common features from 5 different datasets of the "UCI Machine Learning Repository" database obtained through the Kaggle platform. In this study, the success performances of the models created by using fice different classification methods, namely Naive Bayes, Logistics Regression, Random Forest, K-Nearest Neighbors and Support Vector Machines, which are machine learning classification algorithms, were compared. In this study, it is aimed to determine the model that can best predict heart desease by using supervised machine learning algorithms. Possible risk factors that may affect the probability of having heart disease in individuals were examined. One of the main goals of the study is to increase the reliability and predictive accuracy of the classification methods. For this purpose, Bootstrap resampling method has been applied to the data set. The success of each classificassion method that is used, has been compared with the model performance measures on raw data and samples. It has been seen that the most successful model is the Random Forest algorithm.