A NOVEL COVID-19 CLASSIFICATION METHOD BASED ON CURE CLUSTERING

Authors : Bergen Karabulut, Güvenç Arslan, Halil Murat Ünver

Pages : 25-35

View : 67 | Download : 106

Publication Date : 2024-06-30

Article Type : Research Paper

Abstract :COVID-19 is a serious disease that spreads rapidly and affects the world. Alternative methods based on machine learning are recommended to diagnose COVID-19 positive and negative cases cheaper and faster. However, as the data size increases, problems such as space requirement or classification time may arise. KNN (K-nearest neighbor), a simple but effective machine learning method, is widely used in various fields. However, the effectiveness of the KNN algorithm decreases considerably when the sample size is large and the number of features is too large. To solve these problems, it is important to use datasets more effectively and to select meaningful parts of the data. The current study proposes an improved neighborhood-based classification method called CURE-NN and compares its performance with standard NN and KNN algorithms. The proposed CURE-NN method obtains reduced structural information from the data by applying clustering before classification to use the dataset more effectively. The resulting reduced structural information was used as a training set in the classification process. The proposed method was applied to the COVID-19 dataset. With this method, while the classification success is preserved as much as possible compared to the NN and KNN methods, the data used in the test phase is reduced by up to 96%. Experimental results show that the reduced data obtained based on structural information can be used instead of the entire data set. In addition, the method works by using only one neighbor, thus eliminating the need for the K parameter compared to the KNN algorithm.
Keywords : Coronavirus, Covid 19 Diagnosis, K Nearest Neighbor, Cure Clustering, Classification