A Novel Hybrid Machine Learning Approach for the Prediction of Renal Disease

Renal disease is a persistent condition characterized by impaired kidney function, rendering them incapable of their vital roles in waste filtration, hormone production, and blood pressure regulation within the human body. When the kidneys fail to effectively eliminate waste from the bloodstream, it sets off a gradual deterioration of kidney function, ultimately resulting in renal disease. This paper aims to develop a novel hybrid classifier to improve the accuracy of renal disease prediction on a dataset collected from one of the reputed hospitals in India. In this study, a methodology was employed to develop two hybrid classifiers: Random Subspace and Bagging. These hybrid classifiers were constructed using six individual classifiers, namely K-nearest neighbor, Support Vector Machine, Naive Bayes, two-class locally deep Support Vector Machines, two-class Averaged Perceptron, and two-class Bayes Point Machine Algorithms. The performance evaluation of the classifiers was conducted using 10-fold cross-validation, with sensitivity, specificity, accuracy and Kappa as the evaluation metrics. The dataset used in this study was collected from a reputable hospital in India and comprised 650 records with 13 attributes. To address missing values within the dataset, the K-nearest neighbor (K-NN) imputation technique was employed. Hybrid classifiers outperform individual classifiers, particularly in the case of hybrid random subspaces combined with two-class locally deep Support Vector Machines. This hybrid approach achieved a remarkable accuracy of 100%, surpassing other random subspace hybrid classifiers and bagging classifiers. The findings of this study demonstrate the effectiveness of hybrid random subspaces in conjunction with two-class locally deep Support Vector Machine models for enhancing the accuracy of renal disease prediction.