396. A machine learning approach to classify high and low feed efficient or grower mink

Abstract

This study aims at evaluating the feasibility of Machine Learning (ML) algorithms in classification of American mink (Neovison vison) into high/low performing groups for three different traits. Average Daily Gain (ADG), Feed Conversion Ratio (FCR), and Residual Feed Intake (RFI) were calculated for 1,088 mink over a period of 105 days. The 10% high and low feed efficient and grower animals were identified. To classify animals into the high/low feed efficient and grower groups based on the recorded features (initial weight and length at the bigining of the test, sex, age, and color-type), five common ML algorithms including Random Forest, Neural network, K-Nearest Neighbors, Logistic Regression, and Decision Tree were employed. The results showed that Random Forest was the most accurate algorithm in classification of 10% high and low ADG, FCR, and RFI groups with the accuracy of 98, 95, and 100%, respectively.