Score Function Design for Decision Making using Conditional Kullback-Leibler Divergence

This study processes a more discriminative score function or decision measure to address current decision outcomes. Drawing from the information theory, the Kullback-Leibler(KL) divergence is enhanced with additional knowledge; specifically conditional KL divergence. A modified KL divergence is introduced alongside Bayesian theory to inform decision making. The effectiveness of the proposed score function is validated through two examples: the Korean election and car sales problem involving multi-criteria decision-making. Computational examples demonstrate that the proposed score function yields favourable results compared to existing methodologies.