An empirical framework to applying unit testing in operational research

Abstract

Operational Research (OR) models usually deal with uncertain, changing requirements. This leads to a continuous process of adapting and reworking the mathematical code. However, there are scarce mechanisms to control its quality. This is essential to Software Engineering (SE), as it enforces the use of Unit Testing: automatically running tests after any alterations, to assess specific parts of the code. This is done to discover where and how errors are happening, simplifying its correction while evaluating their possible ramifications. This article aims to define how these concepts can be adapted to them, how tests should be used to detect faults and to provide a workflow to use them while developing an OR model. It provides guidelines on what should be tested and what to expect of possible errors and a process to use it.