[2I1-GS-2] Machine learning: Random forest

[2I1-GS-2-05] Predicting Carcass Weight of Wagyu with Regression Analysis and Random Forest

〇Shingo Tsukamoto¹, Takaya Yoshihiro¹, Haruka Ikegami², Tamako Matsuhashi², Kazuya Matsumoto² (1. Wakayama university, 2. Kindai university)

Keywords:Wagyu, correlation coefficient, carcass weight, RandomForest

As Wagyu has been recognized as high-quality beef in the world, livestock farmers of Wagyu are continuing their trials to improve beef quality of their cattle. However, since improvements have been done based on their own experiences and know-how, inheriting their techniques to others is hard. To establish efficient way of raising cattle based on scientific evidence and populating it in Japan would lead to high productivity of Wagyu beef. Wagyu beef is evaluated and priced based on quality. One of the most important quality criteria is carcass weight, which determines the amount of beef meat to be sold. In this study, we propose a new method to predict carcass weight in combination with the traditional regression and the random forest method. We first use regression to predict carcass weight from the initial weight of the cattle, and then learn its errors from a protein expression data set using the random forest. Through evaluation, we found the proposed method outperforms the conventional methods.