Improvement on Data Quality Monitoring Methods Based on Nonparametric Control Charts
In order to monitor the data quality of multivariate data streams,we propose several moni-toring methods.We first introduce some background information of statistical process control,which is necessary to understand this paper.We adopt the Minkowski distance standard,propose 3 EWMA non-parametric control charts based on three kinds of common Minkowski distance.We also propose a fourth EWMA nonparametric control chart based on vector product.In the simulation part,first we compare the performance of the 4 charts when mean shift occur in one data stream under different correlation levels.Then we study how can the smoothing parameter influence the performance of the charts.Later,we apply the 4 charts in the large sample case and propose 5 methods to monitor the data quality of several data streams.The performance of the 5 methods are compared with an existing typical method,in cases when different scales of shift occur.Finally,the advantage of the methods we propose is showed when they are used to monitor real data in a landslide example.
data qualityEWMA control chartsnonparametric control chartsMinkowski distance
万方数据
维普
