IDS-EAPD Project: Intelligent decision support based on explanatory analytics of preference data

Explanatory decision support based on interpretable models constitutes today the greatest challenge for operational research and artificial intelligence. This challenge is called „explainable analytics of preference data„. We take up this challenge by proposing methods involving decision models composed of logical statements called decision rules. They are induced from preference information elicited by decision makers and structured using the dominance-based rough set approach. The decision rules whose general syntax is “if a conjunction of elementary conditions on selected attributes is true, then suggested decision is…” are easily interpretable, and give a clear image of preferences. They identify values that drive users’ decisions – each rule is a scenario of a causal relationship between evaluations on a subset of attributes and a comprehensive judgment.

The overall goal of this project is to develop interactive, highly explanatory decision-aiding methods that allow decision-makers to understand how their preference data were represented in the preference model that led to the recommendation. The main research tasks include:

  • Interactive evolutionary multiobjective optimization driven by decision rules representing preferences expressed both in objective and decision space
  • Interactive multiobjective optimization reinforced by constraints generated by decision rules
  • Construction and explanation of non-compensatory composite indicators using decision rules
  • Interpretation of black-box decision models obtained using neural nets or utility-driven methods in terms of decision rules
  • Consensus reaching in group decision-making using compatible instances of decision makers’ preference models
  • Fuzzy-rough hybridization of granular approximations in view of structuring preference data prior to rule induction
  • Improved algorithms of rule induction, including hierarchical construction of meta-rules and construction of ensemble classifiers composed of diversified decision rules