In industry it is commonly accepted that all temporal models decay given enough time. Statistical process control (SPC) was developed to moniter physical proce

In industry it is commonly accepted that all temporal models decay given enough time. Statistical process control (SPC) was developed to moniter physical processes (such as manufacturing) but can be applied to ML to moniter model output and help determine when retraining is required, or at least alert when a process appears to move outside control. The R qi library generates nice control charts. https://cran.r-project.org/web/packages/qicharts/vignettes/controlcharts.html https://cran.r-project.org/web/packages/qicharts/vignettes/c...