Abstract
Lung disease is generally diagnosed in clinical trials. Several recent epidemiological discoveries have been made through the analysis of relevant biomedical data. Additional studies involving body shape, focused particularly on anthropometrics issues, are usually published in various medical areas. Most of the risk factors have been found; however, the usability of these risk factors (i.e., their predictability or explanation) is less common in practical diagnosis. Compared to previous methods, rule extraction possesses sufficient comprehensibility in medicine. For this reason, a support vector machine (SVM) is used for diagnosis of lung disease in this study. An SVM with prototype method was employed to rule extraction. Next, to enhance the ability of explanation or comprehensible for rule extraction, an active learning-based SVM approach (ALBA) is utilized in this paper. Typical rule generation method and decision trees are used as a benchmark in this paper. The results showed that this approach performed better than a decision tree. In practice, both approaches are reliable and comprehensible. The results of this work demonstrate that both approaches are useful for lung disease prediction and diagnosis.