Multi-criteria Decision-Making Based Classifier Ensemble by Using Prioritized Aggregation Operator

This paper proposes a novel approach for classifier ensemble by employing the concepts of multi-criteria decision-making (MCDM) and aggregation operators. In this framework, a heterogeneous ensemble process has been incorporated where we consider varied set of classifiers to train the model. Each considered classifier is trained on the training data and a score correspondent to it is generated by utilizing the MCDM process. Subsequently, during the training phase, the priority is generated among the classifiers. For the testing phase, these prioritized classifiers are combined using prioritized aggregation operator. The priority order determined during the training phase is used to ensemble the classifiers during the testing phase. The proposed method is tested on UCI benchmark datasets and outperforms existing state-of-the-art methods.