Distribution-free EWMA Chart Based On Online After-sales Review

In the production process,control charts are an effective quality control method.In many industries,Statistical Process Control(SPC)has played a transformative role.If the upward and downward drift of unknown parameters is detected at the same time,the control chart is called a bilateral control chart;A condition in which an unknown parameter drifts up or down is detected is called a unilateral control chart.Traditional control charts are constructed under the assumption of a known process distribution,and this kind of control chart is called a parametric control chart.However,in the actual application process,due to the complexity of the data,the distribution in many processes is difficult to determine,and when the pre-specified parameter distribution is invalid,the results of the parameter control chart will no longer be accurate.In order to solve this problem,many scholars have begun to use nonparametric methods to build nonparametric control charts.Studies have shown that nonparametric charts are generally more stable and effective than parametric charts when the process distribution is unknown.In recent years,due to the rapid development of online shopping,the post-sales online review process has become a key link in product development.With the expansion of information technology in the e-commerce era,more and more customers are publishing reviews of products and services on the Internet.This provides e-commerce businesses with a wealth of user-generated content to measure customer satisfaction,identify potential feedback,and improve e-word-of-mouth.For user-generated content,online reviews are mainly divided into negative and non-negative reviews.Among them,Quality issues,such as failures of products or services,are more likely to be hidden in negative reviews.Therefore,this paper focuses on detecting the abnormal changes of the timebetween-review T and entiment scores S of negative reviews.In practice,the distribution assumptions for S and T may be invalid due to the complexity and variability of online review processes.Therefore,a nonparametric control chart is proposed.Firstly,introduction of a distribution-free exponential weighted moving average(EWMA)control chart for the second stage monitoring,and the statistics S and T are combined using the max-type combination function,and a distribution-free exponential weighted moving average mapping based on Lepage-type test is proposed to detect process scale parameters and location parameters at the same time.This is called a distribution-free exponential-weighted moving average chart based on the Lepage test(Lepage-EWMA).Secondly,the control line is calculated by Monte Carlo simulation stochastic simulation method,and the average running length(ARL),standard deviation of running length(SDRL)and quantile are obtained,and the performance of the distribution-free EWMA chart under control and out of control is analyzed.And comparing the efficiency of different control charts with the expected average run length(EARL)and relative average run length(RARL)of the two performance indicators.Finally,an example is used to illustrate the detection effect of distribution-free Lepage-EWMA chart on detecting location parameters and scale parameters at the same time,and the performance of the chart in practical application is analyzed.