Performance Analysis of Kidney Disease Prediction Using Machine Learning Classification Algorithms and Ensemble Techniques

Chronic kidney disease (CKD) is very dangerous disease and also represented by chronic kidney failure. In CKD, continuous loss of kidneys with time and stops functioning after some time. According to various survey report of World Health Organization, growth rate of CKD patients is increasing rapidly worldwide every year. Kidneys are the preliminary source of our body for blood cleaning by removing waste and it is also highly responsible for regulating salt, potassium and $\mathrm{p H}$ levels in the body. It manages production of red blood cells and also regulates blood pressure in the body by producing hormones. The main aim of this research paper to design predictive model for detection of early CKD that helps patients get medical treatment at the right time and helps to reduce the loss of lives. Artificial Intelligence playing a very demanded role in field of health sector f or prediction of disease and uncountable research studies have been conducted for categorizing CKD using various machine learning classification algorithms. The effectiveness of any algorithm depends upon selecting the specific key features, applying the best feature extraction techniques, and productive classifier models. Proposed model, uses the Pearson correlation coefficient (PCC) method for feature selections and various classification and ensembles algorithms. The accuracy of ensemble models is better than as compared to machine learning models. The accuracy of KNN model is lowest among all model with score ${0 . 6 9 \%}$ and Random Forest model achieved highest accuracy score with value $0,91 \%$. The accuracy score of ensemble boosting model is found to ${0 . 9 8 \%}$. Various accuracy metrics are applied to measure the performance of proposed models like confusion metrics, F1- score, Precision and Recall etc. Among all applied models, the RF classification algorithm and AdaB ensemble methods have achieved high accuracy and linear regression achieved the lowest accuracy.