Trend Forecasts

Some of the items stocked in the inventory have demand patterns where the month-to-month level is gradually increasing (or decreasing) in a steady way, and thereby, a trend forecast model is called. The model has two coefficients, a and b, where a is the intercept and b is the slope. Three such models are described here: trend regression forecasts, trend discount forecasts and trend smoothing forecasts. The trend regression forecast model generates a straight line fit through the most recent N history demands giving equal weight to each history demand. The trend discount forecast model also uses the N most recent history demands, but gives relatively less weight to each older demand. This model is based on a discount parameter, β, that specifies how to apportion the weight to each older demand entry. The trend smoothing forecast model revises the forecast coefficients as each new demand entry becomes available. The model has two parameters, α and β, that are used to revise the trend coefficients, (a, b), at each month. All three of the forecast models generate forecasts that are in fractional numbers and are here called raw forecasts. The forecasts are converted to integers using the cumulative rounding algorithm described in Chap. 2. For latter use, in inventory control, the standard deviation of each of the forecast models is also generated each month. For comparative sake, the coefficient of variation, cov, is also generated each month. Three dampening forecast models are described, to avoid situations when the forecasts are quickly declining to zero and below.