A Model Using Support Vector Machines Recursive Feature Elimination (SVM-RFE) Algorithm to Classify Whether COPD...

Area of Science:

• Pulmonary Medicine
• Medical Informatics

Background:

• Chronic obstructive pulmonary disease (COPD) management requires adherence to Global Initiative for Chronic Obstructive Lung Disease (GOLD) guidelines for optimal patient prognosis.
• Continuous management according to GOLD guidelines is crucial for improving outcomes in COPD patients.

Purpose of the Study:

• To develop and validate a classification model to identify COPD patients managed continuously according to GOLD guidelines over the past year.
• To differentiate between COPD patients receiving consistent GOLD-compliant care and those who are not.

Main Methods:

• A classification model was developed using a prospective cohort (Managed group) and a retrospective cohort (Control group).
• A synthetic minority over-sampling technique (SMOTE) was employed for data balancing.
• Support Vector Machine Recursive Feature Elimination (SVM-RFE) was used for feature selection, with classification performed using LibSVM.

Main Results:

• The final analysis included 15 patients in the Managed group and 191 in the Control group.
• Nine key features were identified, including smoking history, post-bronchodilator FVC, BMI, AECOPD frequency, mMRC score, and post-FEV1 metrics.
• The model demonstrated high performance with a positive predictive value of 66.7%, F1 score of 0.978, and AUC of 0.987.

Conclusions:

• A Support Vector Machine (SVM) classifier combined with SVM-RFE effectively distinguishes between continuously managed and unmanaged COPD patients.
• This model has potential clinical applications to support physician decision-making and improve patient adherence to standard COPD treatments.