Dynamic multi-objective optimization algorithm based decomposition and preference

Most of the existing dynamic multi-objective evolutionary algorithms (DMOEAs) are effective, which focuses on searching for approximation Pareto-optimal front (POF) with well-distributed in handling optimization problems (DMOPs). Nevertheless, real-world scenarios, decision maker (DM) may be only interested a portion corresponding POF (i.e., region interest) different instances, rather than whole POF. Consequently, novel DMOEA based decomposition and preference (DACP) is proposed, incorporates DM into search process tracks subset set (POS) respect to interest (ROI). Due presence dynamics, ROI, defined gives both point neighborhood size, changing time-varying DMOPs. our algorithm moves reference points, located range, around lead evolution population. When change occurs, strategy performed responding current change. Particularly, population will reinitialized according promising direction obtained by letting few solutions evolve independently short time. Comprehensive experiments show that this approach very competitivecompared state-of-the-art methods.