BRSV: A New Hybrid Ensemble Method for Crowd Sourced Data

Thorough comparison of classification and ensemble algorithms is undertaken in this paper in order to study their performance across various datasets. Individual classifiers, as well as ensemble methods, were evaluated with the study itself. Ensemble methods continually outperform single classifiers with regard to accuracy and robustness, depending on the data set. Among the extant approaches, Boosting, Random Forest and Stacking Classifiers performed the best. Voting Classifier, which merges predictions from multiple based models, bears a specific strength towards exploiting single classifier strengths. The conclusion thus proposes that the application of Boosting, Random forest and Stacking in a Voting ensemble framework, BRSV, grants the highest accuracy with an indicated superiority of classification performance and stability by taking advantage of individual classifiers: implying that a voting ensemble integrating these top-performing models would produce the best accuracy.