An Investigation on Multi-Objective Fish Breeding Program Design

Fergus Currie, Yi Mei, Mengjie Zhang, Maren Wellenreuther, Linley Jesson

Abstract: This paper studies the multi-objective fish breeding program design (MO-FBPD), which is an important problem with significant impact in the aquaculture industries. For the first time, we formulated the problem into a multi-objective optimisation problem, including the design of the decision variables, objective functions and constraints. We also developed a simulator for the fish breeding process to facilitate research and analysis. Then, we applied the well-known NSGA-II to solve the MO-FBPD problem, analysed the performance of NSGA-II, and the distributions of the solutions obtained by the algorithm. The results shown are promising. First, it is observed that the breeding program, represented as the top percentage of the fish selected for mating, can generalise to different fish population size. In other words, the trained breeding programs can be applied to future breeding processes with any fish population size. Second, it is observed that the long-term and short-term gains of the fish population are conflicting with each other, and they follow a linear relationship. Finally, if the selection strength is too greedy, it may negatively affect even the short-term gain. This work is a promising preliminary study in this problem domain, which can pave the way for further research in MO-FBPD.