Cow Culling Decisions Adapted for Management with CART
用随机动态规划模型优化牧场淘汰决策，结合CART方法生成决策树，发现双季产犊比传统春季产犊可提高财富7%-10%。
Abstract A stochastic dynamic programming (DP) model of range cow culling decisions incorporating market price uncertainties and dynamics of biological productivity was solved for biannual and annual calving systems. Decision trees were generated from the DP solutions using the Classification and Regression Trees (CART) methodology. The decision trees captured over 99% of the optimal DP returns from both biannual and spring‐only calving. CART culling criteria in conjunction with dual‐season calving increased wealth by 7% compared to optimal DP culling decisions with spring‐only calving, and by 10% compared to a more traditional strategy of culling all open cows.