Monitoring Models And Applications For Complex Changes In Streaming Data

With the progress of science and technology,streaming data gradually appear in every field of life.Therefore,it is necessary to monitor a large amount of data in real time.Due to the characteristics of large volume,fast speed,unidirectional and real-time of the streaming data,it is required to consider both the storage problem and the calculation rate in the monitoring.Since the process distribution has scale shift,location and scale shift,and other shift,this paper proposes three control charts based on these problems.Firstly,this non-parametric control chart that monitors the change of scale parameters is obtained by integrating Klotz test statistics into the change point model.The commonly used threshold value ht is obtained by monte carlo simulation.And the simulation results of the software show that the monitoring performance of this control chart is better.Secondly,by integrating Wilcoxon rank sum test statistics and Klotz test statistics into the change point model,the control chart for monitoring the joint change of location and scale parameters is obtained.The results of computer simulation and experimental comparison show that our control chart has better monitoring performance.Finally,we integrate the likelihood ratio two-sample nonparametric goodness-of-fit test into the change point model to create a new nonparametric control chart to monitor the random changes of distribution.Compared with other nonparametric control charts,our control chart is effective not only in monitoring the change of location shift,scale shift but also in the change of shape shift and so on.And the effect is particularly good when the distribution is scaled down.The algorithm of the three non-parameter self-starting control charts proposed above is also improved.The observation points are separated when monitoringthe streaming data.In the model with moving window,an efficient method for calculating rank is proposed.In addition,this method makes the calculation amount and storage quantity become constant,and enables the streaming data to be monitored online in a timely and efficient manner,thus achieving the purpose of discovering the changes in time and reducing losses.