DL‐HYKIDRECNET: IOT‐Enabled Hybrid Deep Convolutional‐Recurrent Model for Early Detection of Chronic Kidney Disease

BACKGROUND: Chronic kidney disease (CKD) is a chronic and life-threatening disorder that requires early diagnosis and proper management to prevent severe complications. However, missing and imbalanced clinical data, complex nonlinear relationships, and limited generalization across heterogeneous patient records pose challenges to reliable CKD classification. AIMS: This study aims to develop a deep learning-based CKD classification framework that addresses these challenges through systematic preprocessing and nonlinear feature learning. METHODS: The proposed framework uses the publicly available Kaggle CKD dataset containing structured clinical features. The preprocessing stage involves mean imputation for missing values, categorical encoding, MinMax normalization for feature scaling, and the Synthetic Minority Over-sampling Technique (SMOTE) to address class imbalance. The processed data are subsequently passed through deep feature learning layers to capture complex nonlinear relationships among clinical variables. The extracted high-level representations are fed into a fully connected dense classification layer with a sigmoid activation function to generate probability scores for binary CKD prediction. RESULTS: The proposed model achieved an accuracy of 98.8%, precision of 96.8%, recall of 100%, and F1-score of 98.4% on the CKD dataset, demonstrating strong classification performance and consistent behavior across the evaluated data partitions. CONCLUSIONS: The experimental findings indicate the potential of the proposed framework for effective CKD classification. However, its evaluation is limited to retrospective clinical data and does not establish prospective or external clinical validity. Therefore, the framework represents a research-oriented classification approach with potential for future clinical decision-support applications rather than a clinically validated diagnostic system.