Pareto-optimal search-based software engineering, POSBSE : a literature survey

Abstract

Abstract—The Search-Based Software Engineering (SBSE) community is increasingly recognizing the inherit “multiobjectiveness” in Software Engineering problems. The old ways of aggregating all objectives into one may very well be behind us. We perform a well-deserved literature survey of SBSE papers that used multiobjective search to find Pareto-optimal solutions, and we pay special attention to the chosen algorithms, tools, and quality indicators, if any. We conclude that the SBSE field has seen a trend of adopting the Multiobjective Evolutionary Optimization Algorithms (MEOAs) that are widely used in other fields (such as NSGA-II and SPEA2) without much scrutiny into the reason why one algorithm should be preferred over the others. We also find that the majority of published work only tackled two-objective problems (or formulations of problems), leaving much to be desired in terms of exploiting the power of MEOAs to discover solutions to intractable problems characterized by many trade-offs and complex constraints.