Relevancy contemplation in medical data analytics and ranking of feature selection algorithms

Abstract This article performs a detailed data scrutiny on chronic kidney disease (CKD) dataset to select efficient instances and relevant features. Data relevancy is investigated using feature extraction, hybrid outlier detection, handling of missing values. that do not influence the target are removed envelopment analysis enable reduction rows. Column achieved by ranking attributes through selection methodologies, namely, extra‐trees classifier, recursive elimination, chi‐squared test, variance, mutual information. These methodologies ranked via Technique for Order Preference Similarity Ideal Solution (TOPSIS) weight optimization identify optimal features model building from CKD facilitate better prediction while diagnosing severity disease. An ensemble novel similarity‐based classifiers built pruned dataset, results thereafter compared with random forest, AdaBoost, naive Bayes, k‐nearest neighbors, support vector machines. The classifier yields accuracy 98.31% selected extra tree (ETC), which as best TOPSIS.