A new pretreatment approach of eliminating abnormal data in discrete time series
Jun Zhang, Hong Wang · 2005
This paper discusses the approach of eliminating abnormal data in discrete time series and tries to protect the useful information as much as possible. Conventional algorithms result in abnormal data diffusing easily if the cluster abnormal data occur. We introduce the robust weighted average by an iterative calculation procedure combined with conventional approach to eliminate the discrete abnormal data and the cluster abnormal data. The test result shows that the new algorithm reveals more powerful on eliminating abnormal data.