A Hybrid Feature Selection Model for Predicting Chronic Obstructive Pulmonary Disease

Chronic Obstructive Pulmonary Disease (COPD) is characterized by a chronic airflow limitation that is generally progressive and an increased chronic inflammatory response triggered by harmful particles or gases in the airways. In general, symptoms, medical history, clinical examination, and lung ventilation obstruction play a vital role in diagnosis. However, COPD is treatable, even though it is a chronic condition that worsens over time. Furthermore, most patients with COPD can have improved symptom regulation and quality of life with careful treatment and a lower chance of developing other disorders. Therefore, COPD diagnosis is essential in the early stages, as it is treatable and will significantly impact the recovery of a patient's health. With tens of thousands of characteristics in high-dimensional biomedical data, precise and effective identification of the main characteristics in these data might help identify associated disorders. However, biological data frequently contains many irrelevant or duplicated characteristics, which significantly impact later classification accuracy and machine learning efficiency. As a result, for COPD diagnosis, an effective predictive model is needed. This study proposed a hybrid feature selection model to extract the best features from the high-dimensional data. These features are further passed to the classification models to identify the performance of the features on various classification models. According to the experimental data, the suggested hybrid feature selection model could predict COPD with a 95.18 percent accuracy and a Kappa Statistic of 0.9.