Currently, there is a lack of fast and effective methods for identifying redundant constraints in security- constrained economic dispatch (SCED). Moreover, data-driven approaches are prone to false positive (FP) misclassification, which may compromise system security. To address these issues, an improved decision tree (DT) algorithm, namely, an improved classification and regression tree (CART) algorithm combined with an enhanced reduced error pruning (REP) strategy, is proposed for rapid redundant constraints identification. First, the SCED model and the principles of CART are introduced. Second, feature engineering and data preprocessing methods for redundant constraint identification are constructed. Then, an improved CART algorithm incorporating a FP penalty mechanism and a REP strategy based on the FP ratio are proposed. Finally, case studies on the SG-126 system demonstrate that the proposed algorithm can quickly and accurately identify redundant constraints while effectively adapting to extreme FP-sensitive scenarios. The accuracy rate of redundant constraint identification reaches 95.13%, with a FP misclassification rate of zero, and system dispatch time is reduced by 88.22% after eliminating redundant constraints.