A data-mining-based intelligent anti-islanding detection scheme for distributed generation (DG) protection has been presented. The process starts at deriving highly involved features using discrete Fourier transform-based pre-processor at the DG end. The features derived include both positive and negative sequence quantities and related features. Once the features are retrieved, the decision tree is trained to build a data-mining model for identifying the islanding events from non-islanding situations, including disturbances close to islanding conditions. The proposed anti-islanding relay is extensively tested on simulation model and provides encouraging results considering wide variations in operating conditions. Further, the validation is extended on real-time digital simulator module to test the efficacy of the proposed anti-islanding relay. The test results indicate that the proposed anti-islanding relay can reliably detect islanding conditions while meeting the speed criteria.