Comments to Editor-in-Chief regarding the review of CYB-E-2014-07-0677.R1
Carlos R. B. Azevedo, Fernando José Von Zuben · 2015
Abstract—In several applications, a solution must be selectedfrom a set of trade-off alternatives for operating in dynamic andnoisy environments. In this paper, such multi-criteria decisionprocess is handled by anticipating flexible options predicted toimprove the decision maker future freedom of action. A method-ology is then proposed for predicting trade-off sets of maximal hy-pervolume, where a multi-objective metaheuristic was augmentedwith a Kalman filter and a dynamical Dirichlet model for trackingand predicting flexible solutions. The method identified decisionsthat were shown to improve the future hypervolume of trade-off investment portfolio sets for out-of-sample stock data, whencompared to a myopic strategy. Anticipating flexible portfolioswas a superior strategy for smoother changing artificial and real-world scenarios, when compared to always implementing thedecision of median risk and to randomly selecting a portfolio fromthe evolved anticipatory stochastic Pareto frontier, whereas themedian choice strategy performed better for abruptly changingmarkets. Correlations between the portfolio compositions andfuture hypervolume were also observed.Index Terms—Anticipatory learning; Bayesian tracking; multi-objective optimization; hypervolume-based decision making.