Abstract:

Chronic kidney disease (CKD) is still a significant public health issue despite improvements in surgical therapy and medication. Researchers from all around the world have lately created high-performance ways for diagnosis, treatment, and preventative therapy due to the rising prevalence of CKD. These solutions can be more effective if the users are aware of the aspects that are pertinent to the issue. In addition to clinical evaluation, medical data analysis for patients can aid healthcare professionals in the early diagnosis of diseases. Although numerous attempts have been made to improve the effectiveness of the intelligent algorithms that analyse health data to predict CKD, there is still room for improvement. This work intends to give a comprehensive categorization and prediction model for kidneyrelated illnesses. These loss functions, which are employed as a part of a modified Deep Belief Network (DBN) classification approach, are Categorical Cross-entropy activation function and SoftMax activation function, respectively. The proposed model outperforms earlier models since it has a 98.5% accuracy and an 87.5% sensitivity for predicting chronic renal illness (CKD). Modern deep learning algorithms can enhance clinical judgement and enable early prediction of CKD and related stages, according to a data analysis of the available information. This approach might aid in halting the progression of renal disease.

Description:

Restricted Boltzmann Machine, CKD, Contrastive Divergence