Gay, Gregory2020-12-102020-12-102010PROMISE '10 Proceedings of the 6th International Conference on Predictive Models in Software Engineeringhttps://hdl.handle.net/11299/217408Associated research group: Critical Systems Research GroupBackground: Search-based Software Engineering (SBSE) uses a variety of techniques such as evolutionary algorithms or meta-heuristic searches but lacks a standard baseline method. Aims: The KEYS2 algorithm meets the criteria of a baseline. It is fast, stable, easy to understand, and presents results that are competitive with standard techniques. Method: KEYS2 operates on the theory that a small sub-set of variables control the majority of the search space. It uses a greedy search and a Bayesian ranking heuristic to fix the values of these variables, which rapidly forces the search towards stable high-scoring areas. Results: KEYS2 is faster than standard techniques, presents competitive results (assessed with a rank-sum test), and offers stable solutions. Conclusions: KEYS2 is a valid candidate to serve as a baseline technique for SBSE research.Report