To develop an efficient prediction model for chronic kidney disease (CKD) to enhance diagnostic accuracy. Based on the CKD dataset from the UCI database, a case-control design was adopted, with samples divided into a CKD group and a non-CKD group. To address missing values in categorical variables, a combined approach using multiple imputation and K-nearest neighbors (KNN) imputation was employed. All features were normalized using the Min-Max method. Twelve key features were selected using least absolute shrinkage and selection operator (LASSO) regression with five-fold cross-validation. The dataset was split into training and testing sets at a 7∶3 ratio. Five machine learning models—logistic regression, random forest, naive Bayes, K-nearest neighbors, and support vector machine were implemented to diagnose CKD. A secondary meta-learner was constructed using a Stacking ensemble strategy to integrate the strengths of base learners. Results show that the Stacking model outperforms individual models across multiple evaluation metrics, achieving an accuracy of 0.958, an area under the receiver operating characteristic curve (AUROC) of 0.981, and an area under the precision-recall curve (AUPRC) of 0.957. Decision curve analysis further demonstrates that this model provides higher net benefit across a wide range of clinical thresholds. The findings suggest that combining LASSO regression with Stacking ensemble learning can effectively improve CKD prediction performance and offer robust technical support for clinical decision-making.