Application of adaptive neuro-fuzzy inference systems to estimate digestible critical amino acid requirements in young broiler chicks.

Abstract

This study aimed to find the digestible lysine (d.Lys), digestible sulfur amino acids (d.SAA), and digestible threonine (d.Thr) requirements to optimize body weight gain (BWG) and feed conversion ratio (FCR) via adaptive neuro-fuzzy inference systems (ANFIS) using either the Genetic algorithm (ANFIS-GA) or Particle Swarm Optimization algorithm (ANFIS-PSO) in Cobb-500 chicks from 1 to 10\xa0d of age. The range of amino acids was 90 to 115% of the recommendations for male Cobb-500 chicks. The estimated dietary d.Lys, d.SAA, and d.Thr requirements by ANFIS-GA and ANFIS-PSO to optimize BWG were the same and were 12.10, 8.98, and 7.89\xa0g/kg, respectively. The optimum BWG predicted by ANFIS-GA and ANFIS-PSO were 270 and 266\xa0g, respectively for the 1 to 10 d period. The estimated dietary requirements of d.Lys, d.SAA, and d.Thr to minimize FCR at 0.995 by ANFIS-GA were 12.10, 8.98, and 7.89\xa0g/kg, respectively. Although the estimated d.Lys and d.SAA requirements by ANFIS-PSO and ANFIS-GA were identical, the predicted d.Thr requirement by ANFIS-PSO was 0.01\xa0g/kg higher than by ANFIS-GA to minimize FCR at 0.963. Comparison of goodness of fit in term of root mean square error revealed that the ANFIS-GA prediction was more accurate than ANFIS-PSO. This study demonstrates that the hybrid methodology of ANFIS-GA is as an effective and accurate approach to modeling and optimizing nutrient requirements.