A Novel Method to Predict Chronic Kidney Disease using Optimized Deep Learning Algorithm

Chronic Kidney Disease (CKD), a significant global health issue and has a major impact on a vast number of people. In the early stages, it is impossible to identify the disease’s symptoms. Very few people are aware of this illness and are able to foresee the symptoms earlier. The disorder is considered as potential risk factor. As a result, a potent deep learning framework is used in this paper to reduce the risk of acquiring these diseases as a way to prevent them. The machine learning algorithms used in early disease prediction, has been found to be computationally expensive, frequently overfit, and underperform in terms of accuracy since they must examine the large amount of clinical data until the model converges. Therefore, in this paper three novel work has been proposed an efficient novel hybrid feature selection strategy RFITLO algorithm is used to find the optimal features that gives the major contribution is classifying the CKD disease. Then two proposed classification algorithms namely Enhanced Multi-Layer Perceptron (HW-MLP) and Optimized Multi-Layer Perceptron (PKD-OMLP) are used in prediction model to capture the complex patterns and optimize the learning algorithm to predict the CKD at prior stage from the data gathered in Kaggle, Real Time and UCI Machine Learning Repository Dataset. In order to measure the classifications of disease, performance measures including accuracy, precision and recall are analyzed. The experimental findings show that the PKD-OMLP strategy produces better results than the proposed HW-MLP and other conventional approaches like Support Vector Machine (SVM), Linear Regression (LR) and Multi-Layer Perceptron (MLP). Among the preceding four models PKD-OMLP renders the best outcome as per its performance level producing a high accuracy of 94.89% on testing Real Time (RT) CKD dataset comparatively with other datasets such as Kaggle and UCI repository. Therefore, these proposed algorithms can support clinicians to enable secured and accurate classification of disease with better accuracy and has the potential to be used effectively in the early detection of chronic kidney disease.