Chronic kidney disease (CKD), which has a high incidence and frequently gets diagnosed late, leads to decreased treatment efficacy and rapid progression to renal failure. Traditional diagnostic methods of serum creatinine and eGFR tests are not as sensitive or fast in detecting CKD in patients who could benefit. However, by the time these abnormalities are detected by routine screening, they are typically at a later stage. The paper proposes a unique predictive analytic methodology that uses machine learning (ML) algorithms to improve the early identification of CKD. The proposed strategy uses a hybrid framework to connect the Random Forest (RF), Support Vector Machine (SVM), and Gradient Boosting algorithms (GB). The goal is to examine data on patient demographics, clinical biomarkers, and past medical records. Data collecting, pre-processing approaches, including feature selection utilizing PCA (Principal Component Analysis) and RFE (Recursive Feature Elimination), as well as model cross-validation training. Performance measurements show a significant improvement over current techniques, with accuracy of 94.5%, sensitivity of 92.3%, specificity of 95.8%, and AUC-ROC more than 70%. It represents a significant step toward improving early detection and forecast accuracy. The approach is more effective and efficient than existing ones; when integrated into a clinical decision support system, the ensemble model ensures improved patient outcomes through faster intervention and increased diagnostic accuracy.