Synthetic datasets for concept drift detection purposes

In order to know about the performance of a drift detector by measuring the different detection metrics, we need to know beforehand where a real drift occurs. This is only possible with synthetic datasets. The scikit-multiflow framework allows generating several kinds of synthetic data to simulate the occurrence of drifts.

为通过测量不同检测指标以评估漂移检测器（drift detector）的性能，我们需要预先知晓真实漂移的发生位置。此类信息仅可通过合成数据集获取。scikit-multiflow框架支持生成多种类型的合成数据，用以模拟漂移的发生。