Weighted Majority Voting by Statistical Performance Analysis on Ensemble Multiclassifier

Ensemble classifier method uses several base classifiers to predict a new test instance, while weighted majority voting has a scheme providing different weight values using several measurement parameters. However, the determination of the appropriate weight value to obtain an adequate ensemble model is a critical issue. This study, therefore, proposed a novel weighted majority voting scheme involving five base classifiers based on ensemble learning, including Random Forest, Decision Tree (C.45), Gradient Boosting Machine, XGBosst, and Bagging. The weighting scheme was formulated by analyzing the base classifier performance measured from the parameters of accuracy, recall, precision, and F Measure. The experiments were conducted using public datasets and umbilical cord data owned and the results showed the proposed method has the ability to improve performance in comparison with the base classifier and methods from previous studies with the best recorded in umbilical cord dataset with an average accuracy of 86.1%, a precision of 86%, a recall of 86%, and an F measure of 86%.