Random Forest Models Of 305-Days Milk Yield For Holstein Cows In Bulgaria

In the field of stock breeding the research of the factors which influence the highest levels of the productivity (for example milk yield) is of significant importance both for the selection of animals and for determining the conditions of their raising. This article researches the influence of 11 linear type traits of the Holstein Friesian breed of cows on the adjusted milk yield for 305 days. The purpose of this work is to demonstrate the possibilities of the method Random Forest (RF) method for building models with high enough statistical significance to determine the main linear type traits of the animals influencing the milk yield. From a sample of 97 cattle observations from 4 farms in Bulgaria two RF models were built to study the dependence of 305 days Milk yield of the Holstein Friesian breed in terms of 12 independent variables - 11 linear type traits and the farm where the animals are raised. The first model is based on 11 independent variables for linear type traits, and the second is based on the same 11 independent variables as the farm has been added. Both models describe about 95% of the data and identify and agree on the main linear type traits that affect the milk yield - udder width, chest width, hock development and locomotion. The models are compared with the published CART models of the same data set.