What is a Seasonality Forecast?
In time series data, seasonality refers to the presence of variations which occur at certain regular intervals either on a weekly basis, monthly basis, or even quarterly (but never up to a year). Various factors may cause seasonality – like a vacation, weather, and holidays. They comprise repetitive, periodic, and generally regular patterns that are predictable in a time series level.
How to Conduct a Seasonality Prediction
In a time series, seasonality fluctuations can be differentiated from cyclical patterns. Cyclical patterns take place when the data shows rises and falls which are not in a fixed timeframe. These non-seasonal fluctuations are generally a result of economic conditions and usually relate to the business cycle. Also, cyclical patterns usually span a minimum of two years.
Seasonal variation is calculated in terms of an index known as a seasonal index. It refers to an average which can be used for comparing a real observation in relation to what it would be if no seasonal variation existed.
Each time series has an index value attached to it within a year. This means that twelve different seasonal indices exist when considering monthly data, one for each of the months. The methods below utilize seasonal indices for measuring seasonal variations of time-series data.
- Method of simple averages
- Ratio to trend method
- Ratio-to-moving-average method
- Link relatives method
Using the ratio-to-moving-average method too measure seasonal variation provides an index for measuring the seasonal variation degree in a time series.
The index is based on a mean of 1oo, having the seasonality degree measured by variations far from the base. For instance, if the hotel rentals in a winter resort are observed, it is then discovered that the winter quarter index is 124.
The value, 124, shows that 124% of the average quarterly rental takes place in winter. If for the whole of the previous year, the hotel management records 1,436 rentals, then 359 = (1436/4) would be the average quarterly rental. Since the winter-quarter is 123, the number of winter rentals is estimated as follows: 359 (124/100) = 445; In this case, the average quarterly rental is 359 while the winter-quarter index is 124. The seasonal winter-quarter rental is 445. This method is also referred to as percentage moving average method. Here, the original data values in the time-series are shown as moving average percentages. The step, as well as, tabulations are given below:
Steps:
- The first step is to find the twelve monthly or four quarterly moving averages of the real data values in the time-series.
- Next is expressing each time-series original data as a percentage of the equivalent centered moving average values gotten in the first step. This means that, in a multiplicative time-series model, we get (Original data values) / (Trend values) 100 = (T C S I) / (TC) 100 = (S I) 100. This means that the ratio-to-moving average represents the components that are seasonal and inconsistent.
- Arrange the percentages based on the months or quarters of given years. Then find the averages of overall months or quarters of the given years.
- If the summation of these indices isnt 1,200, (or 400 for quarterly figures), multiply then by a correction factor = 1,200 / (monthly indices sum). Otherwise, the twelve monthly averages would be seen as seasonal indices.