model, is an effective methodology for decision making from vague, uncertain or imprecise data with different classification costs. Over the past decades, several measures for attribute reduction in the decision-theoretic rough set model are presented. The monotonicity of those measures is, however, not always satisfied. In this paper, a novel monotonic measure is introduced for attribute reduction in the decision-theoretic rough set model. Based on the concept of the maximum inclusion degree and maximum decision, maximum decision entropy is firstly proposed, and the definitions of the positive, boundary and negative region preservation reducts are then provided by using the measure of the maximum decision entropy. Theoretically, it is proved that the proposed measure is monotonic as adding or deleting the condition attribute. Additionally, a maximum decision entropy-based attribute reduction algorithm is developed for the decision-theoretic rough set model, which maximizes the relevance of the reduct to the class attribute and also minimizes the redundancy of the condition attributes within the reduct. The experimental results on synthetic as well as real data