Advice - Forecasting
General
You can assume normal distribution for your count data
In practice, this rarely matters provided our counts are sufficiently large. If the minimum number of customers is at least 100, then the difference between a continuous sample space
and the discrete sample space {100,101,102,…} has no perceivable effect on our forecasts. However, if our data contains small counts (0,1,2,…), then we need to use forecasting methods that are more appropriate for a sample space of non-negative integers.
- Forecasting: Principles and Practice
Don't use MAPE!
Check Time series metric
Add the correct number of dummy variables
Many beginners will try to add a seventh dummy variable for the seventh category. This is known as the “dummy variable trap”, because it will cause the regression to fail. There will be one too many parameters to estimate when an intercept is also included. The general rule is to use one fewer dummy variables than categories. So for quarterly data, use three dummy variables; for monthly data, use 11 dummy variables; and for daily data, use six dummy variables, and so on.
Combine your forecasts!
The results have been virtually unanimous: combining multiple forecasts leads to increased forecast accuracy. In many cases one can make dramatic performance improvements by simply averaging the forecasts.
- ...benefit of combining forecasts. That is, find two or more valid forecasting methods and then calculate averages of their forecasts. For example, make forecasts by using different regression models, and then combine the forecasts. This is especially effective when the methods, models, and data differ substantially. Combining forecasts has reduced errors from about 10% to 58% (depending on the conditions) compared to the average errors of the uncombined individual forecasts (Graefe, et al 2011). (From Armstrong 2012 ) (note that this is Ensemble modeling)
Simple forecast methods are better
- Simple forecasts seem to outdo complicated forecasts by alot. More accurate and also understandable (from Green and Armstrong 2015)
ARIMA
Difference the seasonality first.
- If there is seasonality in your data, difference the seasonality first! It could be that the data will become stationary (see Stationarity). If you do the first differencing, there is still seasonality (from Forecasting: Principles and Practice)
Use dummy variables to account for holidays
- If you want to account for holidays in your time-series, you can add a dummy variable (0 for non-holiday, 1 for holiday) in an ARIMAX.