An application of Boosted Classification and Regression Trees (CART) in agricultural ergonomics

Abstract

Classification and Regression Trees (CART) is a decision tree based approach widely used for classification when dependent variable is categorical and for regression when dependent variable is continuous. The advantages of this approach are that they are non-parametric, data-driven, can handle outliers, suitable when there are interactions between the independent variables. In spite of several advantages, some of the drawbacks of CART approach are low model accuracy, high prediction variance and model overfitting. Boosted CART is one of the popular approaches for dealing with this problem in which multiple trees are trained sequentially based on information from previously grown trees. In this study, two approaches conventional CARTand Boosted CART have been compared for classification problem. Empirical results based on simulation as well as real datasets uncover that Boosted CART performed best in terms of classification accuracy over conventional CART.