Machine Learning for Enhanced COPD Diagnosis: A Comparative Analysis of Classification Algorithms.

Area of Science:

• Pulmonary Medicine
• Medical Informatics
• Artificial Intelligence

Background:

• Chronic Obstructive Pulmonary Disease (COPD) is a leading cause of death in the U.S.
• COPD is a chronic inflammatory lung disease causing airflow obstruction and symptoms like breathlessness and cough.
• COPD patients face increased risks for comorbidities such as heart disease and lung cancer.

Purpose of the Study:

• To evaluate the effectiveness of various machine learning (ML) classifiers for diagnosing COPD.
• To compare the performance of different ML algorithms in identifying COPD from patient data.

Main Methods:

• The study utilized a dataset of 1603 patients referred for pulmonary function tests.
• Ten machine learning classifiers were applied: Logistic Regression, Gradient Boosting Classifier, Support Vector Machine, Gaussian Naïve Bayes, Random Forest Classifier, K-Nearest Neighbors Classifier, Decision Tree, and Artificial Neural Network.
• Model performance was assessed using accuracy, F-score, and ROC values.

Main Results:

• The Random Forest Classifier (RFC) demonstrated the highest accuracy, achieving 82.06% in training and 70.47% in testing.
• RFC also achieved the maximum F-score and an ROC value of 0.82.
• Overall model accuracies ranged from 67.81% to 82.06% (training) and 66.73% to 71.46% (testing).

Conclusions:

• Machine learning models show significant potential for COPD diagnosis.
• The Random Forest Classifier emerged as a highly effective tool for identifying COPD.
• These findings align with previous research, supporting the utility of ML in respiratory disease diagnostics.