Hybrid Method of Selection Features to Improve Performance of Covid-19 Classification

Abstract

In this study, to improve the performance of the classification algorithm using the Hybrid feature selection method on covid-19 data, this study utilizes the SBS Filtering and Wrapper Technique, aiming to reduce the initial feature sub-space dimensions N (total features) to K (features). best). From the whole process of testing the Hybrid method by combining filtering and wrapper techniques, it can be concluded that from a total of 8974 features, after entering the filtering process, 184, then after applying the wrapper technique to 170 selected features, performance evaluation was carried out and obtained SVM performance results with data The big one is with an accuracy of 83.8% and testing on KNN by testing the parameter value K = 5, getting an accuracy result of 79.5%, the classification of the K value is determined by the researcher. The overall precision comparison is KNN with a precision value of 32.6% and SVM with a precision value of 87.6%, recall with a KNN result of 14.2%, and SVM of 20.1%, a comparison of F1-Score KNN of 17.3% and SVM of 27.5%.