SBSR Solution Evaluation: Methods and Challenges Classification

Abstract

Search-based software engineering has made substantial progress in recent decades. It is an approach to software engineering in which search-based optimization algorithms are used to solve software engineering problems. Search-based software refactoring (SBSR) is used to improve the quality and maintainability of the software by finding a proper sequence of refactorings using search-based optimization algorithms. Fitness functions utilized to evaluate refactoring solutions in search-based algorithms play a vital role. Indeed, the more accurate and more effective the fitness function is, the more reliable the provided solutions will be. In this paper, we plan to accurately assess this area from the perspective of metrics used in fitness functions to evaluate solutions. To this end, first we propose a classification of metrics used for solution evaluation. Then, the challenges of each category are investigated. Understanding the challenges and ways to handle them can lead to an important comparison and assessment of the presented approaches. It will also direct researchers to accurately recognizing and improving existing approaches in the future.