Random Subset Feature Selection and Classification of Lung Sound

Abstract

The lung sounds produced by a human convey valuable information about the health of the respiratory system, and these signals are complex in nature. In this paper, a study was conducted to find the importance of feature selection from these signals for the purpose of classification. Feature selection is performed using two different approaches: RSFS and SFS. The experiment was conducted on a dataset of 85 samples using the (SVM, KNN, and Naïve Bayes) classifiers. The computational results obtained are promising, and the proposed feature selection techniques show better performances in terms of Precision, Recall, and F-Measures.