Definicja time series model
One of three categories of methods used for forecasting. The other two are causal models and judgmental models. Time series models develop forecasts by assessing the patterns and trends of past sales. The key determinant in the selection of a time series model is the pattern of previous sales data. The pattern of sales is the profile or general premise that future sales will mimic past sales - the pattern of past sales needs to be identified and incorporated into the forecast. Once the pattern is identified, the forecaster can select the time series model best suited to that particular pattern. For example, if there is a seasonal influence in past sales (e.g. sales are consistently highest in October and April), then a forecasting model that compensates for seasonality should be used (i.e. classical decomposition or Winters method). If past sales have small fluctuations with no major pattern or trend, then some type of smoothing model (moving average or exponential smoothing) might be best.
Time series techniques all have the common characteristic that they are endogenous methods. This means that a time series model looks only at the patterns of the history of actual sales (or a series of sales through time, hence the term time series). If these patterns can be identified and projected into the future, you can use them to produce a forecast. Time series models are the most commonly used methods for forecasting.
Słownik i definicje SAPa na T.