Hybrid Genetic Algorithm and Simulated Annealing for The Selection of Web-Based Beef Cattle Feed Composition

Abstract

The nutritional requirements for the fattening process in each beef cattle differ according to body weight and body weight gain targets. Inappropriate feed composition can be detrimental to breeders because the bodyweight gain target is not achieved, and improper expenditure of feed funds. Genetic algorithms can be used to search for feed composition solutions, where the nutrients produced are close to the nutrients needed by beef cattle. Genetic algorithms have several disadvantages, one of which often occurs premature convergence, where genetic operators cannot produce offspring better than their parents. Premature convergence on genetic algorithms can be overcome by hybridizing local search algorithms, one of which is Simulated Annealing. Simulated Annealing is a local search algorithm that functions as a counterweight to genetic algorithms, where genetic algorithms are able to explore global areas, while simulated Annealing is able to exploit local areas. Comparative testing of hybrid genetic algorithm and simulated Annealing with a simple genetic algorithm shows that the fitness value of the hybridization method is better than the simple genetic algorithm. The best fitness of the hybridization method is 0.15934987829563, and the best fitness is a simple genetic algorithm of 0.15869962195529. The hybridization method produces better fitness because of the role of simulated Annealing in exploiting individuals on genetic algorithms so that the composition of feed solutions can be closer to the optimal solution.