Advanced Content Master's Series
Division: All - Tools/Skills
Statistical process control for autocorrelated processes have been addressed in previous publications using the exponentially weighted moving average (EWMA) one-step-ahead forecast or simple auto-regressive integrated moving average (ARIMA) models. The time series model forecasts the motion in the mean and an individuals control chart is plotted of the residuals to detect assignable causes. Failure to account for the autocorrelation will produce limits that are too narrow, resulting in excessive false alarms or limits that are too wide, resulting in misses.
This session introduces recent developments in time series forecasting that improve upon the simple models using modern techniques such as automatic model selection. While these methods are computationally complex, they are easy to use and validate. They have been tested in forecast competitions using thousands of data sets demonstrating very good forecast accuracy over a wide range of sample sizes and presence of seasonality in the data. While obviously beneficial in forecast applications such as customer demand and sales, we will apply the model residuals for control chart purposes. A better forecast will mean a better control chart in the presence of autocorrelated data.