Supply_Chain_7.ppt

Supply Chain Management--Suqian Zha * Forecasting Supply Chain Requirements * Forecasting Supply Chain Requirements Forecasting Supply Chain Requirements (2) Classic Time Series Decomposition A class of forecasting models that has been useful over the years is that of time series decomposition. These methods include spectral analysis, classic time series analysis, and Fourier series analysis. Classic time series decomposition analysis is discussed here mainly because of its mathematical simplicity and its popularity. And because more elegant methods have not offered increased accuracy. Classic time series decomposition forecasting is built on the philosophy that a historical sales pattern can be decomposed into four categories: trend, seasonal variation, cyclical variation, and residual, or random vaviation. Classic time series analysis combines each type of sales variation in the following way: F=T*S*C*R F=demand forecast T=trend level S=seasonal index C=cyclical index R=residual index In practice, the model is often reduced to only trend and seasonal components. This is done because a well-specified model has a residual index value(R) of 1.0 and thus does not affect the forecast, and because it is difficult in many cases to decompose cyclical variation from random variation. Treating the cyclical index (C) as equal to 1.0 is because the model is usually updated when new data become available. The mathematical expression for a linear trend line is T=a+bt, the coefficients are found by ∑Dt(t)-N(D)(t) b= ∑t2-Nt2 a= D-bt N= the number of observations used in the development of the trend line Dt=the actual demand in time period t D=average demand for N time periods t=average of t over N time periods St=Dt/Tt Ft=(Tt)(St-L) St= seasonal index in time period t Tt=trend value determined from T=a+bt Ft=the forecasted demand in time period t L=number of periods in the seasonal cycle Example: A manufacture of young wome