FORECAST.ETS: How do I add the seasonality

Patsarat 

You're on the right track! Here's how to address seasonality in your FORECAST.ETS model with a time-based unit of "Year" and a peak period of 3 months (October, November, December):

1. Seasonality Period:

• Yes, you should add a seasonality of 3 to the FORECAST.ETS function. This indicates that the data exhibits a recurring pattern every 3 units (in this case, months within a year).

2. Unit Compatibility:

• While the time-based unit in your model is "Year," the seasonality period is defined in months. This doesn't necessarily pose a problem. The FORECAST.ETS function can handle different units for time-based data and seasonality as long as they are convertible.

Here's how to incorporate seasonality in your FORECAST.ETS formula:

Explanation:

• values: This is the range of historical data you want to forecast.
• timeline: This is the range of future dates for which you want predictions.
• [data_completion]: (Optional) This argument specifies how to handle missing data points.
• [aggregation]: (Optional) This argument defines how to aggregate multiple data points with the same timestamp.
• 3: This is the seasonality parameter set to 3, indicating a recurring pattern every 3 months.

Additional considerations:

• Ensure your data covers at least multiple cycles of the seasonality period (ideally 2-3 years) for the model to learn the seasonal pattern effectively.
• You can experiment with different seasonality values (e.g., 4, 6) to see which one provides the best forecasting accuracy based on your data.
• If your data exhibits trends along with seasonality, consider using the FORECAST.ETS.LINEAR or FORECAST.ETS.EXP functions, which can account for both trends and seasonality.

By incorporating seasonality correctly, you can improve the accuracy of your forecasts and capture the periodic fluctuations in your data.