Transforming Machine Learning Models to Decision Support Systems

Abstract

We present a systematic approach for transforming machine learning (ML) models into decision support systems (DSS) suitable for real-world deployment. While ML models often achieve strong predictive performance, prediction alone is frequently insufficient in operational settings where decisions must comply with business rules, policy constraints, and requirements for stakeholder interpretability and justification. Moreover, many practical problems require actionable recommendations that go beyond patterns observed in historical data, introducing a decision layer not inherently captured by standard ML models. To address this gap, we propose an efficient and straightforward transformation process that augments ML models with decision-support capabilities. The approach integrates domain constraints, policy alignment, and explicit decision logic while preserving the predictive strengths of the original models. Empirical evidence suggests that this transformation can be achieved with minimal loss in predictive performance. We argue that such DSS-enhanced models provide a more appropriate framework for decision-making in complex organizational environments, where correctness, transparency, and actionability are as critical as accuracy. Our work positions DSS not as an alternative to ML, but as a necessary extension that bridges the gap between prediction and decision, particularly in business contexts where accountable and policy-aligned decisions are essential.

Sandro Radovanović

Assistant Professor at University of Belgrade

My research interests include machine learning, development and design of decision support systems, decision theory, and fairness and justice concepts in algorithmic decision making.