Weighted pointwise prediction method for dynamic multiobjective optimization Ali Ahrari a,⇑ , Saber Elsayed a , Ruhul Sarker a , Daryl Essam a , Carlos A. Coello Coello b a School of Engineering and Information Technology, University of New South Wales, ACT, Australia b CINVESTAV-IPN, Departamento de Computación, Mexico, Mexico article info Article history: Received 30 September 2019 Received in revised form 16 June 2020 Accepted 7 August 2020 Available online 24 August 2020 Keywords: Evolutionary algorithm Dynamic problem Multi-objective optimization Reinitialization Continuous optimization abstract Prediction methods are useful tools for dynamic multiobjective optimization (DMO), espe- cially if the changes roughly follow some patterns. Multi-model prediction methods, in particular, may capture different types of change patterns; however, they should address two issues. First, they should define a similarity measure that can correctly find the corre- sponding Pareto-optimal solutions in two successive time steps. Second, they should be reasonably robust to input errors. This study introduces a new information-sharing strat- egy to improve the robustness of multi-model prediction methods in which each predic- tion model utilizes some information from the individual models of adjacent solutions. An adaptive scheme based on the relative distribution of population members is also pro- posed to utilize this information properly. The efficacy of this strategy in improving the robustness of the multi-model prediction method is demonstrated. Furthermore, this study introduces a similarity metric and thoroughly analyzes it alongside some of the commonly used similarity metrics for DMO. A weighted pointwise prediction method (WPPM) for DMO is then developed using the formulated information-sharing strategy and the pro- posed variable-based similarity metric. WPPM is compared with other well-known predic- tion methods on the CEC2018 test suite for DMO, with the numerical results revealing the superiority of WPPM. Ó 2020 Elsevier Inc. All rights reserved. 1. Introduction Dynamic multiobjective optimization (DMO) problems refer to a class of problems in which multiple objectives are opti- mized when a problem’s landscape changes over time. This change may originate from a change in the formulation or the number of objective functions [8], variables [26] or constraints [2]. DMO methods have gained increasing popularity in the last decade since many practical problems are subject to dynamic changes and conflicting objectives. Several previous stud- ies have investigated the merits of DMO in varied applications, such as: design of control systems [21], mission planning [5], wastewater treatment [18], and vehicle routing [17]. Dynamic optimization is motivated by the fact that in many practical dynamic problems, changes are generally not rad- ical [6], and thus, the problem landscape after a change is correlated to the previous landscape. Furthermore, the changes in a problem may follow a pattern which can be exploited [41,19]. One example is a steadily changing environment in which for https://doi.org/10.1016/j.ins.2020.08.015 0020-0255/Ó 2020 Elsevier Inc. All rights reserved. ⇑ Corresponding author. E-mail addresses: a.ahrari@unsw.edu.au (A. Ahrari), s.elsayed@unsw.edu.au (S. Elsayed), r.sarker@unsw.edu.au (R. Sarker), d.essam@unsw.edu.au (D. Essam), ccoello@cs.cinvestav.mx (C.A. Coello Coello). Information Sciences 546 (2021) 349–367 Contents lists available at ScienceDirect Information Sciences journal homepage: www.elsevier.com/locate/ins