Sınıflandırma temelli veri madenciliği teknikleri kullanılarak koroner kalp hastalığı (KKH) tanısı

Coronary heart disease (CHD) has attracted the most attention around the world because it leads to death. These days, data mining in many fields, including commercial fields and medical fields, where medical fields are the most productive of large data on a continuous basis, and which must find different ways to extract information, may be important in predicting the spread of this disease. We have designed a system to help the diagnosis of CHD with better reduction of costs and time required for the process by using a programing language with data mining classification techniques. These algorithms produced good results and high accuracy. We applied our study to various CHD datasets. We obtained the best accuracy at 99% through the use of the Random Forest (RF) algorithm with Hungarian two classes. With Cleveland, we obtained 94% accuracy using the same algorithm while the better accuracy with the same dataset in the previous study was 58% when using the SVM algorithm. Moreover, with the Hungarian five class dataset, we obtained 99% as the best accuracy using random Forest Classifier algorithm rather than the accuracy that was achieved with this dataset in previous work, which was close to 67% using the SVM algorithm. In addition, we obtained 88% as a better accuracy using the AdaBoost classifier with the Hungarian data set and 87% accuracy using the Logistic Regression classifier with the heart.csv dataset. With the Switzerland dataset, we had 95% as the best accuracy using Random Forest and 91% best accuracy with the Long-Beach dataset using the same classifier. Finally, with the Switzerland dataset, we achieved a 78% better accuracy using the AdaBoost and Logistic Regression classifier. With Long-Beach, we had 80% using the AdaBoost classifier and 76% xii using the Logistic Regression classifier. Also with the heart.csv dataset, we achieved 87% best accuracy using the Logistic Regression classifier and 86% accuracy when using the AdaBoost classifier. We used a train test split and preprocessing for the CHD dataset in this study and processed the missing values that were found with attributes with a less complicated system. This process differs significantly from previous study is proposed results and accuracy for this purpose with the same CHD dataset.