Reusing Relational Queries for Intuitive Decision Optimization

Decision optimization is used in many applications such as those for finding the best course of action in emergencies. However, optimization solutions require considerable mathematical expertise and effort to generate effective models. On the other hand, reporting applications over databases are more intuitive and have long been established using the mature database query technology. A decision optimization problem can be viewed as an "inverse" of the reporting problem. For example, a report may tell the total cost of a certain supply chain given the various sourcing and transportation options used; the corresponding optimization problem can be to select among all possible sourcing and transportation options to minimize the total cost. Reusing existing reporting queries for decision optimization will achieve the dual goals of taking advantage of past investments and of making decision optimization more intuitive. To realize these goals, this paper addresses two related technical issues with a decision guidance query language (DGQL) framework. The first is to annotate existing queries to precisely express the optimization semantics, and the second is to translate the annotated queries into equivalent mathematical programming (MP) formulation that can be solved efficiently. This paper presents the decision queries with examples, provides formal syntax and semantics to DGQL, describes an implementation method through a reduction to MP formulation. Finally, the paper illustrates via experiments on a prototype system that the optimization tasks done with DGQL compete squarely with expertly generated MP models.