Multi-Objective Search-based Requirements Selection and Optimisation

Most software product developments are iterative and incremental processes that are seldom completed in a single release. It is critical but challenging to select the requirements from a large number of candidates for the achievement of overall busi- ness goal and the goals of multiple stakeholders, each of whom may have competing and often conflicting priorities. This thesis argues that search-based techniques can be applied to the optimisation problem during the requirements selection and analysis phase for release planning problem. Search-based techniques offer significant advantages; they can be used to seek robust, scalable solutions, to investigate trade-offs, to yield insight and to provide feedback explaining choices to the decision maker. In the thesis, a Search-based Requirements Selection and Optimisation Framework is proposed that includes search spaces, representations, solution processes and em- pirical studies. This framework formulates the requirements selection problem as an optimisation problem and allows multi-objective search-based techniques to be used in order to provide optimal or near optimal solutions and to find a suitable balance between priorities in different contexts. The thesis reports the results of experiments using different multi-objective evo- lutionary optimisation algorithms with real world data sets as well as synthetic data sets in three studies of the applications of this framework: Value/Cost Trade- off in Requirements Selection, Requirements Interaction Management and Multi- Stakeholder Requirements Analysis and Optimisation. Empirical validation includes a statistical analysis of the performance of the algorithms as well as simple graph- ical methods to visualise the discovered solutions in the multi-dimensional solution space. Though these visualisations are not novel in themselves, the thesis is the first to use them for visualisation of requirements optimisation spaces.