This paper presents predictive analytics for chronic kidney disease (CKD) based on clinical features using a powerful deep learning-based model. We tested performance of four existing best deep learning algorithms, convolutional neural network (CNN), recurrent neural network (RNN), multilayer perceptron (MLP), and long short-term memory (LSTM). Our research, based on a large set of pre-processed clinical variables, has indicated that MLP model had a superb testing accuracy of 93.94%, thus it is the most accurate model. Interpretability of model was enhanced by the use of SHapley Additive exPlanation (SHAP) which is an interpretable machine learning methodology. TimeToEventMonths, eGFRBase-line, ACEIARB, CreatnineBaseline and HbA1C were the most significant clinical factors used to predict the model, based on SHAP analysis. The interpretability created by SHAP substantially enhances clinical decision-making by offering a reasonable and coherent interpretation of results projected. The findings provide additional evidence that interpretable methods, combined with advanced deep learning approaches, can contribute to the prediction of CKD. © 2025 IEEE.