A hybrid of decomposition and domination based evolutionary algorithm for multi-objective software next release problem
Xinye Cai, Ou Wei · 2013
In software industry, one common problem that the companies face is to decide what requirements should be implemented in the next release of the software. In a more realistic perspective, multiple objectives, such as cost and customer satisfaction, need to be considered when making the decision; thus the multi-objective formulations of the NRP become increasingly popular. This paper studies various multiobjective evolutionary algorithms(MOEAs) to address multi-objective NRP(MONRP). A novel multi-objective algorithm, MOEA/DD, is proposed to obtain trade-off solutions when deciding which requirements to be implemented in MONRP. The proposed MOEA/DD addresses several important issues of decomposition based MOEAs in context of MONRP by combining it with desirable feature of domination based MOEAs. A density based mechanism is proposed to switch between decomposition and domination archives when constructing the subpopulation of subproblems. Our experimental results suggest the proposed approach outperforms the state-of-art domination or decomposition based multi-objective MOEAs.