Objective Based on machine learning algorithms, a prediction model for major adverse cardiovascular events (MACE) during hospitalization in patients with chronic obstructive pulmonary disease (COPD) is constructed, providing a basis for early identification of patients at high risk of MACE. Methods A retrospective collection of COPD patients hospitalized in the Department of Respiratory and Critical Care Medicine of a tertiary general hospital in Luzhou from January 2024 to October 2025 was conducted as the study subjects. The occurrence of MACE during hospitalization was used as the outcome variable. Missing data were handled using multiple imputation by chained equations (MICE). All candidate variables were directly entered into LASSO regression for variable selection. Various machine learning models were used to build prediction models, and the performance of each model was compared using accuracy, precision/positive predictive value, recall/sensitivity, F1 score, area under the receiver operating characteristic curve (AUC), calibration curve, and decision curve analysis. Results A total of 541 hospitalized COPD patients were included, among whom 181 experienced MACE, with an incidence rate of 33.5%. Compared with naive Bayes, Extreme gradient boosting, random forest, and light gradient boosting models, the logistic regression model had better overall performance and was easier to interpret. Its AUC was 0.836 (95%CI 0.771–0.902), accuracy was 0.778, precision/positive predictive value was 0.688, recall/sensitivity was 0.611, F1 score was 0.647, and specificity was 0.861. Multivariate logistic regression analysis showed that statistically significant independent factors included gender, prothrombin time, uric acid, self-care ability, N terminal pro B type natriuretic peptide, smoking history, and stroke. Conclusions The incidence of MACE during hospitalization is relatively high in COPD patients. The logistic regression model predicts the risk of MACE in hospitalized COPD patients quite well and is clinically applicable, with model coefficients directly indicating the direction and magnitude of each predictor's effect.