DOE-Synthetic Extremely Imbalanced Binary Classification Dataset

The Synthetic Imbalanced Classification Dataset was developed to support empirical investigations of binary classification performance under extreme class imbalance. Its generation process follows a statistically rigorous Design of Experiments (DOE) framework, enabling controlled manipulation of six primary generative factors: feature dimensionality, sample size, imbalance ratio, response function, decision threshold, and stochastic error variability. These factors systematically regulate the complexity, separability, and noise characteristics of the synthetic samples, ensuring reproducible instantiations of challenging classification regimes across varying difficulty levels.Each dataset instance contains a structured feature matrix, a binary response variable aligned with the specified imbalance ratio, and detailed metadata encoding the experimental conditions used during synthesis. The resource is optimized for benchmarking classifier robustness in settings where minority class detection is operationally critical.