主题: |
On Surveillance of High-Dimensional Datastreams |
类型: |
学术报告
|
主办方: |
|
报告人: |
邹长亮教授(南开大学) |
日期: |
11月9日(周1)下午4:00-5:00 |
地点: |
知新楼B-1238 |
内容: |
报告题目:On Surveillance of High-DimensionalDatastreams 报告人:邹长亮教授(南开大学) 报告时间:11月9日(周1)下午4:00-5:00 报告地点:知新楼B-1238 报告摘要: Monitoring high-dimensional data streams hasbecome increasingly important for real-time detection of abnormal activities inmany statistical process control (SPC) applications. This talk consists of twoparts, which are the outlier identification (Phase I) and sequential detection(Phase II). In the first part, I will introduce an outlier detection procedurefor high-dimensional data. The method is to replace the classical minimumcovariance determinant estimator with a high-breakdown minimum diagonal productestimator. The cutoff value is obtained through the asymptotic distribution ofthe distance, which enables us to control the type I error and deliver robustoutlier detection. In the second part, we propose a test statistic which isbased on the "divide-and-conquer" strategy, and integrate thisstatistic into the multivariate EWMA charting scheme for on-line detection. Thekey idea is to combine many T-square statistics calculated on low-dimensionalsub-vectors. The proposed procedure is computation- and storage-efficient. Thecontrol limit is obtained through the asymptotic distribution of the teststatistic under some mild conditions on the dependence structure. Both(asymptotically) theoretical analysis and numerical results show that theproposed method behaves well in high-dimensional data. 欢迎各位老师和同学参加! |