A novel method using Covid-19 dataset and machine learning algorithms FOR THE MOST ACCURATE DIAGNOSIS that can be...

Pandemics and many other diseases threaten human life, health and quality of life by affecting many aspects. For this reason, the medical diagnosis to be applied for any disease is important in terms of the most accurate determination by the doctors and the appropriate treatment for the determined diagnosis. The COVID-19 pandemic that started in China in December 2019 spread all over the world in a short time. Researchers have begun to do different studies to make the most accurate diagnosis of COVID-19. Due to the rapid spread of COVID-19, doctors in the health sector of many countries were also caught off guard. Machine Learning Algorithms (MLAs) are of great importance in the development of computer-aided early and accurate diagnosis systems in today's medical field, as they greatly assist doctors in the medical diagnosis process. In this study, a method was proposed for the most accurate diagnosis of COVID-19 patients using the COVID-19 image data. Images were first standardized and features extracted using RGB values of 800x800 images, and these features were used in train and test processes for MLAs. 5 different MLAs were used in experimental studies using statistical measurements (k Nearest Neighbor (k-NN), Decision Tree (DT), Multinominal Logistic Regression (MLR), Naive Bayes (NB) and Support Vector Machine (SVM)). A method was proposed that automatically finds the highest classification success that these algorithms can achieve. In experimental studies, the following accuracy rates were obtained in train operations for MLAs, respectively; 1, 1, 1, 0.69565, 0.92753. Accuracy results in test operations were obtained as follows; 0.85714, 0.79591, 0.91836, 0.61224, 0.89795. After the application of the proposed method, the test success rate for MLR increased from 0.91 to 0.98. As a result of applying the proposed algorithm, more accurate results were obtained. The results obtained were given in the experimental studies section in detail. The results obtained proved to be very promising. According to the results, it was seen that the proposed method could be used effectively in future studies.