|
Abstract:
|
Often, time series data are intermittent (discontinuous or interrupted). Very common in business and economics, intermittent time series are mostly constant with occasional departures from the base value. Commonly used forecasting techniques are interval-based time series methods such as exponential smoothing methods (ESM) and Box-Jenkins methods (ARIMA). These methods are meant to forecast the future values with respect to future time periods. Since the most likely future value is the base value, these models are inadequate when compared to the naïve model of simply the base value. In contrast, intermittent demand methods (IDM) forecast the future average demand per period, which is more appropriate especially for many inventory control systems. Additionally, IDMs are useful for forecasting time series data that are hierarchical in nature. The aggregation can be temporal or spatial.
|
