Knowledge-Driven Decision Optimization for Non-Experts

Abstract

Decision optimization is a pressing need of small and large enterprises, who are trying to succeed in a competitive and rapidly changing marketplace. The creation of decision optimization models typically requires modeling the problem in a manner that can be solved by optimization engines such as IBM CPLEX. While high-level modeling languages exist for many solvers, their use requires significant optimization expertise, due to the need to model the problem in terms of constraining variable values, and to create models that can be solved efficiently. This severely limits the widespread use of optimization. In this demo, we present an alternative approach, which enables users who are not optimization experts to write programs that verify that a given solution satisfies the problem constraints and compute the value of the objective function. These programs are written in popular languages such as Python; our technology, demonstrated here, analyzes such specifications and produces models for an optimization solver. We believe that this approach will enable the wider community of data scientists and developers to model optimization problems, enabling a much more widespread use of optimization.