Linear Programming Initialization Method of Evolution Strategies for Beef Cattle Feed Optimization

Abstract

The biggest expense in cattle fattening is feed costs. Minimize expenses without ruling out the nutrients needed by the cattle is necessary. This study uses Evolution Strategies (ES) to optimize the beef cattle feed. However, the performance of the conventional ES takes a long time to obtain the optimal solution. This may be caused by an error in initial population which is a significant assignment in evolutionary algorithms since it can influence the convergence speed and the final solution quality. Random initialization is the most commonly utilized strategy to initialize the population. This paper proposes a novel initialization approach which utilizes Linear Programming (LP) to produce initial population. The experiments show that the ES (µ/r+) with LP surpassed the LP and ES (µ/r+). The fitness value obtained from ES (µ/r+) with LP was 13.12156 and the price was 763.8331 with zero penalties. The proposed method produce the highest fitness value with the lowest price and the lowest penalty, thus prove the usefulness of LP as the initial population of ES.