Analysis of data mining for classification of Obstructive Sleep Apnea in chronic obstructive pulmonary disease patients

Abstract

Chronic obstructive pulmonary disease (COPD) is one of diseases that could cause a problem of significant concomitant chronic disease which increases morbidity and mortality. COPD is characterized by airflow resistance in the airways caused by airway abnormalities or anatomical abnormalities of the lungs or a combination of both. One complication that can occur in patients with COPD is lack of oxygen intake at night. This situation will be further aggravated if people with COPD also suffer from Obstructive Sleep Apnea (OSA) sleep disorders. In this study, we used Information Gain feature selection to determine which features that affect the risk of OSA in COPD patients. After the feature selection process was completed, we used the Random Forest method to classify who has a high risk and who has a low risk of developing OSA in COPD patients. The sample in this study consist of 111 COPD patients with 34 features who were hospitalized in X Hospital during March 2018 to May 2018. From an observational result, after we choose 5 %, 10 %, 20 %, 30 %, 40 %, 50 %, 60 %, 70 %, 82 %, and 100 % best features of total features, the best accuracy is obtained by 10 % of best features total features (4 best features) i.e. 85.71 % with sensitivity and specificity are 71.43 % and 92.86 % respectively. The feature with the highest ranking is waist size.