An Ensemble Learning Approach for Chronic Kidney Disease Prediction Using Different Machine Learning Algorithms with Correlation Based Feature Selection

Chronic Kidney Disease (CKD), also known as Chronic Renal Disease is considered one of the biggest reasons acting behind deaths in adults all over the globe and the number is escalating throughout the years. At its final stages, treatment of CKD becomes much exorbitant. Machine learning algorithms, for their capabilities to learn from experience, can play a vital role in predicting CKD in its early stages. In this paper, we apply machine learning to predict CKD on the basis of clinical data obtained from the UCI machine learning repository. The dataset has a significant amount of missing values which is handled using K-Nearest Neighbours imputer. The imbalanced dataset has been balanced using Synthetic Minority Oversampling Technique (SMOTE). A Correlation Based Feature Selection (CBFS) and Principal Component Analysis (PCA) is used for feature selection. Later, the dataset is divided into 80% for training, 10% for validation and 10% for testing. Five renowned supervised learning algorithms namely K-Nearest Neighbours (KNN), Support Vector Machine (SVM), Gaussian Naive Bayes, Decision Tree, Logistic Regression and an Ensemble learning algorithms are used to achieve the prediction. Among these, the ensemble learning algorithm proves to be superior than others on the dataset obtained by CBFS, acquiring an accuracy, precision, recall, and f1-score of 97.41%, 99.52%, 95.27% and 97.33% respectively.