Interpretable systems based on evidential prospect theory for decision-making

Dempster Shafer Theory is known for its capability of modelling information uncertainty by considering the powerset of decision alternatives. Studies in literature propose numerous solutions to resolve the open issues in DST like basic probabilities computation, and conflicting evidence combination. However, there is no widely accepted method so far which can resolve both the issues simultaneously. This work presents a Decision Support System based on descriptive decision-making model which attempts to resolve both the issues, and provides interpretable knowledge about the decision space. The proposed DSS considers triangular fuzzy number to compute basic probabilities, and multi-criteria decision-making methods, instead of DS combination rule, to assign fusion probabilities. The decision alternatives are ranked based on fusion probabilities by an optimal MCDM method, and gain-loss values from prospect theory. Experimental analysis is performed on ten benchmark datasets from various domains. A comprehensive comparison of results with traditional approaches and with recent research works are presented. It can be inferred that VIKOR method has assigned high fusion probabilities, but its prediction accuracy is less compared to TOPSIS; moreover variations in the gain-loss values corresponding to fusion probabilities is observed due to various decision-maker's attitudes towards risk. An optimal MCDM method, TOPSIS, is chosen based on its performance and prospect values. The approaches and outcomes of this work can be used to develop explainable decision support systems for various applications.