Evolutionary algorithms with average crossover and power heuristics for aquaculture diet formulation

The aquaculture farming industry is one of the most important industries in Malaysia since it generates income to economic growth and produces main source of food for the nation. One of the pillars in aquaculture farming industries is formulation of food for the animal, which is also known as feed mix or diet formulation. However, the feed component in the aquaculture industry incurs the most expensive operational cost, and has drawn many studies regarding diet formulation. The lack of studies
involving modelling approaches had motivated to embark on diet formulation, which searches for the best combination of feed ingredients while satisfying nutritional requirements at a minimum cost. Hence, this thesis investigates a potential approach of Evolutionary Algorithm (EA) to propose a diet formulation solution for
aquaculture farming, specifically the shrimp. In order to obtain a good combination of ingredients in the feed, a filtering heuristics known as Power Heuristics was introduced in the initialization stage of the EA methodology. This methodology was capableof filtering certain unwanted ingredients which could lead to potential poor solutions. The success of the proposed EA also relies on a new selection and
crossover operators that have improved the overall performance of the solutions. Hence, three main EA model variants were constructed with new initialization mechanism, diverse selection and crossover operators, whereby the proposed EAPH-RWS-Avg Model emerged as the most effective in producing a good solution with the minimum penalty value. The newly proposed model is efficient and able to adapt to changes in the parameters, thus assists relevant users in managing the shrimp diet formulation issues, especially using local ingredients. Moreover, this diet formulation strategy provides user preference elements to choose from a range of
preferred ingredients and the preferred total ingredient weights.