【摘要】 流数据的海量、无限、分布动态变化且不均衡等特征使得对流数据的在线异常检测成为当前一个研究热点。分析了异常数据的少而不同且更容易通过随机空间的分割而孤立出来的特征,基于在线集成学习理论,提出了一种基于隔离森林的在线流数据异常检测算法。在4个UCI标准数据集上的实验结果表明提出的方法有效。更多还原

【Abstract】 Streaming data has some unique characteristics such as massive,unlimited generation,dynanmic variation distribution and unbalanced data distribution and so on,which make the anomaly detection for streaming data become a research hot spot.An obvious characteristic of anomalous sample is"few and different"compared to these normal data,which makes it more easily to be isolated than normal data by the partition of stochastic space.In this paper,a novel online anomaly detection method for streaming data was proposed based on the isolation principle and online ensemble learning theory.Experiments conducted on four UCI datasets demonstrate the effectiveness of the proposed method.更多还原