Use of machine learning methods in classification of respiratory system diseases

The diagnosis and classification of asthma, COPD and pneumonia, which are the most common respiratory diseases, has been vital importance. Recently, there has been great interest for processing health data due to this importance. Machine learning algorithms have an important place for this purpose. The focus of previous studies has been to obtain the most predictive algorithm for diagnosing asthma, COPD and pneumonia diseases. However, the determination of the most significant variables has received little attention. The principal objective of this study is to present the most predictive machine learning algorithms with variable importance results in details. A comprehensive comparison study based on the machine learning algorithms and variable importance evaluations have been carried out by considering different performance criteria including accuracy, precision, recall, Kappa statistics, F-measure, ROC curve and AUC value. The findings have been indicated that Random forest (with CART learner), C5.0 for asthma disease and SVM (with non-linear kernel), GBM for COPD disease and Bagging ve Random Forest (with CART learner) for pneumonia have been found as the best algorithms. Also, GBM is the best tested algorithm for the classification of all groups together. Considering the variable significance, the most important variables were FEV3, FVC, FEV1, MEF50 according to the classification results of asthma cases between healthy group and diseases. MEF50, FVC, FEV3, PIF and FEV1 in COPD and FVC, FEV3, MEF50 and FEV1 in pneumonia are remarkable variables. Almost all medical variables are of great importance in the classification of all groups. The rank of important variables has been supported by correlation analysis and statistical significance tests. Additionally, ROC curves and AUC values have provided similar and supportive results as visually. Consequently, this study will be of value to practitioners and researchers studying on expert systems on health sciences and machine learning applications. Keywords: asthma, COPD, pneumonia, disease prediction, machine learning, feature selection.