A multi-period forecasting method for intermittent demand based on non-overlapping temporal aggregation

Intermittent demand is common in practice. Inventory managers expect to know the future demand of each period over a long horizon for inventory control. Most existing studies on forecasting intermittent demand can only forecast one future period. Though the multi-period demands can be forecasted by some existing estimators, of which the trend coefficients are static, which may not align with the trend of future periods. In this paper, a multi-period forecasting method based on non-overlapping temporal aggregation is proposed. Firstly, the framework of the proposed method can forecast the future demand of each period over a long forecasting horizon by the patterns of non-overlapping aggregated time series with different aggregation level. Secondly, three principles are extracted, and then three strategies are proposed based on the principles. The three strategies, abridging or extending the original time series, can cope with the circumstance that the quantity of original historical periods cannot be divided by the aggregation level. Thirdly, four properties of the proposed method are discussed and illustrated in theory. Then, seven basic estimators and two accuracy measurement methods are introduced. The results of numerical and empirical investigations illustrate that the method is practical and efficient.