This study deals with a short-term forecasting of the sales quantity of two cosmetic products-emollient creams and lotions-produced by a company in Korea.
The annual sales volume of the two products(among fifteen products handled by the company) amounts to about 30 percent of the total sales volume, and each sales has a seasonal trend.
ARIMA models are developed through three-stage iterative procedures, I.E., identification, estimation and diagnostic checking. In specific, autocorrelation and partial autocorrelation functions of each time series are used to specify various alternatives, to which the nonlinear least squares method is applied to estimate parameters. Then, residual autocorrelation and partial autocorrelation functions are used for diagnostic checking on each alternative model for its validity. If one of the alternatives is specified as an adequate one from the diagnostic checking procedure, then it is chosen as the short-term forecasting model.
Through the above procedures forcasting models for the two products are selected. All the required computations are performed with a computer package for Box-Jenkins Methodology available in Korean Advance Institute of Science \& Technology.