
Part 2. MM4XL Tools > 1. Strategic Tools > Forecast Manager > 1. How To Run Forecast Manager > Page 2: Data Attributes Forecast ManagerPage 2: Data AttributesClicking on the tab Data Attributes opens the panel below. The frame Quality of Forecast is an important one because it hosts the Measure of best fit command, which is the parameter Forecast Manager uses for ranking the quality of fitted curves. There are 5 different measures of fit to choose among: MAD, MAPE, RMSE, MSE, and R squared. You will find more about reliability measures later in the section Technicalities. By now, it is worth saying that these coefficients determine the discrepancy between actual and fitted values. The best curve is chosen among the hundreds or thousands of curves Forecast Manager tests.
The drop down box Moving periods is used for running the three moving average methods available. Each of the buttons in the frame Cockpit opens a new page where users can refine their forecast. The radio buttons Select all and Deselect all allow checking at once all options available under each button. The Reports button shows controls for printing the Forecast chart with actual and fitted curves placed between the boundaries of a confidence interval (CI). Fitted values lying outside the boundary are highlighted on chart with either a green (above CI) or red marker (below CI). The Full accuracy report prints for each selected model all the coefficients computed for measuring the quality of fit. Beside the 5 measures of fit mentioned above also Theils U and DurbinWatson indices are shown. The Confidence interval can be set at the desired percentage level between 1% and 100% (of each forecasted value). The Special events chart shows the forecast error between limit boundaries set by default at 1 and 2 standard deviations. Values crossing the limit are highlighted as being affected by some sort of special effect presumably produced by marketing actions, such as promotional campaigns, for instance.
The button Seasonal Cycle allows modeling seasonal cycles. When checked the Print seasonal indices checkbox prints on sheet the seasonal coefficients computed on the actual values. The selection in the picture below refers to a quarterly series (4 seasonal periods) beginning with the first quarter.
The Indices button opens a page with one single option called Reliability & Accuracy Measures that prints a summary table showing all major coefficients needed for evaluating the quality of fit. Beside the coefficients introduced above, turningpoints are counted, both missed and false signals, and the consistency of performance is measured in terms of values below and above confidence interval.
The Error buttonopens the frame below. The Error summary checkbox prints a table showing, for each item in the series, error terms in value, percent, and cumulative percent. The Cusum chart option, whose default limits are set at 2SD (2 standard deviations), prints a chart for monitoring the correct functioning of the forecast algorithm. The user can change it to 1 or 3 standard deviations.
The button Solver shows the panel where to set the Solver parameters. However, you hardly have to interact with these values. Change them only if you are confident with Solver, otherwise it is not required. The default settings solve most optimizations models without need for changes. In case Solver cannot reach a solution according to the default parameters the user is asked to proceed or stop the optimization algorithm. The algorithm, however, stops for one iteration only. Given it takes several iterations for reaching an optimal solution, Solver may well start again even after you chose to stop.
List of constraints to page Data Attributes:
