Chronic kidney disease (CKD) is a prevalent disease that presents a variety of challenges, including soaring costs for intervention, urgency, and more importantly, difficulty in early detection of the disease. It is a non-communicable disease that affects 10–15% of the global population. It depicts a medical condition that harms the kidneys and affects a person’s overall health. Early and accurate detection of the stages of CKD is vital to minimize the impact of patients’ health complications such as hypertension, anemia (low blood count), mineral bone disorder, poor nutritional health, acid base abnormalities, and neurological complications with timely intervention through appropriate medications. In the field of medical science, Machine Learning (ML) techniques have become valuable tools and play a significant role in disease prediction. The aim of this study was to develop and validate a predictive model for the diagnosis of chronic diseases. A dataset of chronic kidney disease with 400 samples was obtained from Kaggle. The current study used a prediction-based method that helps in detecting and diagnosing CKD patients, which enables a fast and accurate decision-making process at an early stage. Although various diagnostic methods are available, this study utilizes machine learning because of its high accuracy. In this study, we used a hybrid technique to build the proposed model. In our proposed model, we used the Pearson correlation for feature selection. In this step, the best models were selected based on critical literature analysis. In the second step, a combination of these models was used in the proposed hybrid model. The objective of this study was to evaluate the best machine learning classification techniques in terms of accuracy in predicting CKD. And also provides a solution for over fitting or imbalanced data and achieves the highest accuracy with batter performance by using re-sampling method k-fold validation which evaluates the performance of Machine Learning. And for handling class imbalance in dataset using the SMOTE technique. Here eight machine learning classifiers, LR, RF, SVM, GBM, KNN, NB, ANN and K-Means Clustering, were used for analysis, and the stacking ensemble method was used to enhance the model performance. Logistic regression produced the best results, with an accuracy of 98.80% that’s why we are using logistic regression as our Meta model. Finally, we achieved the highest accuracy of 100% using the stacking ensemble approach.