Chronic Obstructive Pulmonary Disease (COPD) is a progressive lung condition that causes breathing difficulties and affects millions worldwide. Acute exacerbations of COPD (AECOPD), a sudden worsening episode requiring emergency care or hospitalization, are a leading driver of patient morbidity and hospital costs, making early risk detection a clinical priority. Inspired by the previous work on the COPDGene cohort, a large dataset containing clinical information from over 10,000 COPD patients, this study investigates whether machine learning (ML) and deep learning (DL) approaches can improve upon earlier classification benchmarks. The input features to the models consist of approximately 300 clinical variables per patient, including demographic information, medical history, and symptom questionnaires. The models are designed to predict exacerbation frequency, classified into seven categories ranging from zero to six or more exacerbations per year. We apply a range of classical machine learning methods, including Support Vector Machines and Gradient Boosted Decision Trees, alongside contemporary deep neural network architectures used to predict patient exacerbation frequency. After identifying the most informative clinical variables, we systematically compare model performance to determine which approaches best capture the relationships between patient characteristics and exacerbation risk. We hypothesize that deep learning architectures will demonstrate improved accuracy in capturing exacerbation risk, owing to their ability to handle more abstract representations of the input data. This work aims to establish updated performance benchmarks and identify promising directions for computational approaches to COPD exacerbation prediction.