Trading between classification accuracy and information in production

There is a tradeoff between the accuracy of a classification model and the amount of information it provides. Increase in the amount of information in a decision often comes at the expense of the decision accuracy. For example, discretization of a continuous target variable such as a production tool work in process (WIP) using more levels increases information but also raises the errors in WIP discretization. An information measure (IM) that trades between the two, using the mutual information between predictions and true decisions, is proposed. The superiority of IM over other performance measures is manifested in various scenarios. In addition, an unsupervised, IM-based discretization strategy is suggested. This strategy determines the number and positions of the discretization splits to increase the amount of information in the discretization while minimizing the error severity. The strategy is applied to the discretization of WIP in a chain of tools of a production FAB.