Search based software engineering

Abstract

Consider the following questions, which are posed by software engineers on a daily basis: 1. What is the smallest set of test cases that will cover all statements in this program? 2. What is the best way to organise classes and methods for this OO design? 3. What is the set of requirements that balances software development cost and customer satisfaction? Whilst these questions seem to be addressing different problems, they do have some notable commonalities. Firstly, they form part of a large set of soft- ware engineering problems which can each be solved by a multitude of potential solutions. That is to say that if one were to ask the above questions to x equally competent engineers, one would likely get back x different yet correct solutions. Secondly, this class of problems is usually tasked with balancing a number of competing constraints. A typical example here is maximising customer satisfaction whilst keeping development costs low. Finally, whilst there is typically no perfect answer (and indeed no precise rules for computing the best solution), good solutions can be recognised. When problems with similar characteristics were encountered in disciplines other than software engineering, they were solved with a large degree of success using search-based techniques. It was this realisation that gave rise to the field of search based software engineering.