A Machine Learning Approach to Predict Meat Production Factors in Japanese Black Cattle Fattening

Abstract

Thanks to its unique characteristics, such as the level of marbling, Japanese Wagyu beef is considered one of the highest quality meats in the world. These characteristics are the result of a wide variety of factors including genetics, production systems, diets, breeding techniques, and environmental conditions. However, the farmer's profit is strongly related to the cost of production (especially the cost of feeding) and to the quality score that each animal obtains at the slaughterhouse. For this reason, this study aimed to build, test, and optimize a machine learning algorithm for the prediction of individual carcass traits, which could be used by farmers as a support tool. To achieve this result, data on environmental conditions, behavior, and blood composition were obtained from 44 Japanese black cattle raised for beef production. The obtained databases were then optimized through two techniques of feature selection: genetic algorithm and correlation analysis. For each of the resultant databases, a neural network was built and tested. The results showed a promising ability of machine learning algorithms to predict carcass traits with good accuracy even with a small sample size. Especially, the genetic algorithm optimized database resulted in the best solution, obtaining a higher accuracy (R2=0.34) with respect to the complete database (R2=0.27) and the correlation optimized database (R2=0.09). This study provides a first step forward in the use of machine learning techniques for the optimization of Wagyu beef production.

Keywords: Japanese Wagyu, Feature Selection, Precision livestock farming, Decision Support System, meat quality