A Proposed Model for Prediction of COVID-19 Depend on K-Nearest Neighbors Classifier:Iraq Case Study

Jaleel, R. A.; Burhan, I. M.; Jalookh, A. M.

The 2019 coronavirus (COVID-19) disease has caused devastation all over the world and is underway in great efforts to control it. In the present time, one of the important interests of decision makers and managers in almost all kinds of hospitals to adopt discovery plans correctly patient's status (Positive, Negative). Accordingly, managements of hospitals become mindful about the status of these patients. In particular, to ensure adequate care at the appropriate time for the patient. The interest of the role of data mining has increased from here that its aim is to discover information from large volumes of data. In this paper, K-Nearest Neighbors (K-NN) algorithm was applied to construct a model of classification for predicting patients' status using actual dataset gathered from the many Iraqi hospitals through a questionnaire prepared and distributed for 290 patients. A large number of features are included in the medical data sets. When the data sets contain noisy features, the performance of the classifier is reduced. To solve this problem, select the subset feature. The selection of features enhances accuracy. Thus to get a highly accurate model and identifying the most effective features that affect the status, two experiments were executed, first experiment implemented without using K-NN and second experiment implemented based on CorrelationAttributeEval, GainRatio AttributeEval, and ReliefF AttributeEval algorithms that are running in WEKA tool for enabling decision makers in hospitals to predict and enhance the status of their patients. Finally to observe the effectiveness of using feature selections, using k-fold cross-validation we assess a proposed model in the two experiments. We found that prediction scores such as accuracy, balance accuracy, precision, recall, true negative rate, and F1 score of K-NN with ReliefF AttributeEval is better than other algorithm of feature selection. © 2021 IEEE.