Diagnosing Chronic Kidney Disease Using Hybrid Machine Learning Techniques

The chronic kidney failure is a serious health issue and if not detected and treated at the early stages, it can be verydeadly. Hence the major objective of this paper is to develop a reliable machine learning model which predicts the CKD with ahigh accuracy rate. The CKD data set is downloaded from the famous UCI ML repository but it suffers from a lot of missingvalues. To handle the missing values KNN Imputation is used. Feature selection is also performed with the help of informationgain as the dataset is huge and hence the cost of modelling can be very high. Various other pre-processing steps like labelencoding and Min-max normalization is performed to attain a clean dataset. After pre-processing, various ML algorithms likelogistic regression, naive bayes, artificial neural network and random forest are applied and their performances are comparedwith the help of various performance metrics. A hybrid of Random Forest and Adaboost algorithm is proposed and it achievesa better accuracy when compared to the other individual component models and hence it can be proved that the proposedhybrid model is much better and accurate in diagnosing CKD.