Positive And Negative Signals Diagnosis Of High Dimensional Data Stream Based On MDR Control

High-dimensional data streams often appear in social life.Multi-variable statistical process control(SPC)tends to be more ambitious,which includes online monitoring and diagnosis.Most of previous studies have paid attention to Online-Monitoring.However.when monitoring complex systems involving high-dimensional data flows,it's vital to di-agnose the abnormal data flow with signals A signal refers to the the data flow which shift when OC occurs.The positive shift is called a positive signal,the other is called a negative signal.Some scholars have studied the diagnosis of signal flow,but cannot diagnose the shift direction.The method proposed in this paper is a high-dimensional data flow fault diagnosis method based on MDR control,which can not only accurately diagnose these signal flows,but also diagnose whether the signal is positive or negative.Specifically,we transform the high-dimensional data flow diagnostic problem into a bi-directional multi-ple hypothesis detection problem,and provide an diagnostic program for the diagnosis of positive and negative signals.The proposed procedure overcame the limitations of tradi-tional diagnostic procedures by controlling MDR at the significance level and minimizing the expected false positive number(EFP).In this paper,we prove theoretically that the proposed program is asymptotically efficient and optimal,and compare the performance of the proposed method with the existing method by numerical simulation,and apply the proposed method to the semiconductor data.