+1 following this thread.
2 of 2 people found this helpful
The forecasting model in V8 is an exponential smoothing model. This basically allows us to calculate weighted averages and then subsequently apply higher weights to the points in time nearest the date we are trying to forecast while applying lower weights to those points farther in the past.
Regarding the quality, the forecast quality will increase as you have more points upon which Tableau can apply the forecasting model. With a few points, you will either get no forecast at all or a low quality forecast.
In addition to the minute amount of information I have given here, the V8 documentation will have a section that is solely based on Forecasting so that everybody can understand and use the forecasts as this is one of the lynch pins of our products' design: analytics for everyone.
The major difference on the positive of Tableau is that anybody can use Tableau to create an exponential smoothing forecast with their data. This is often not true with stats packages. However, within tools built specifically for statistics, there are going to be more aspects of your forecasting models that can be controlled I suspect.
Do not, however, lose sight of the fact that the iteration of forecasting in V8 is the first.
Hope this helps,
Do you have any familiarity with ARIMA forecasting Models . I tried forecasting with Tablaeu and trend and seasonality is not good enough .for my application ..for others I am sure it is. What I have is a weekly pattern , that has a very small trend ocer the months , but the weekly pattern is approximately sinusoidal , but its amplitude changes from week to week , so effectively I have two sinusoids , one inside the other , the inside one changing daily ,ie a wekly sinusoid, and the outside one changing sinusoidally , with a month period. with a seasonaly period yearly , but trending up slowly from year to year.ts actually the prepaid charging behaviour in a cellular network...
Please advise ...maybe I should do it in R , and the try get the formual and put it in Tableau , where I just tune the ceoffiecients as the forcast proceeds