CKD Prediction using Mixed Dataset Concept and Deep Learning: A Combined Approach for Accurate Diagnosis

Chronic kidney disease (CKD) is a significant health concern worldwide, requiring early detection and intervention for improved patient outcomes. In this study, a novel approach is proposed, that combines two diverse datasets, obtained from Kaggle and UCI repositories, to create a balanced and comprehensive dataset for CKD prediction. Unlike conventional methods, feature selection is refrained from in order to retain the richness of information present in the datasets. The combined dataset undergoes undersampling to address class imbalance, followed by training a Feedforward Neural Network (FNN) model for accurate prediction. This experiment reveals promising results, with an accuracy of 98.88% and a low loss of 0.0259. Additionally, a case study is presented to illustrate the practical application of the model by predicting CKD status based on a specific row index. This study demonstrates the potential of utilizing mixed datasets and deep learning techniques for robust CKD prediction, thereby facilitating early diagnosis and intervention.