Ensembling of non-linear svm models with partial least square for diabetes prediction

Abstract

This paper focuses on the improved prediction of diabetes over the very famous Pima Indians dataset. This work focuses on the ensembled result of Non-linear support vector machines (SVM) aggregated with partial least square classifier (PLS). The idea behind this is to get the advantage of kernel transformations from the Non-Linear methods of SVM and dimensionality reduction from PLS. This unique combination makes the ensembled classifier efficient which can be observed after comparing it with the previous classifiers. So this method is also compared with the leading classifiers like decision tree, neural network, linear SVM and also with the ensembled models of these classifiers by applying majority voting. In all the cases, the proposed method is doing better than the rest of the methods. © Springer Nature Singapore Pte Ltd. 2020.