Demand forecast of a product family on different hierarchies

• Main Author: Ong, Li Fong.
• Other Authors: School of Mechanical and Aerospace Engineering
• Format: Final Year Project
• Language: English
• Published: 2010
• Subjects: DRNTU::Engineering::Industrial engineering::Information systems
• Online Access: http://hdl.handle.net/10356/40305
• Institution: Nanyang Technological University

Demand forecasting is inevitable for companies to adopt in today’s context as it determines the demand for the future, bringing in advantageous benefits to a company. In this study, two ways of forecasting were evaluated, namely, to forecast directly on the component level and the other was to forecast on the module level which would be then converted to the component level. Also, ARIMA and Croston methodologies were being evaluated. To assist the research process, two (ARIMA and Croston) methodologies, out of the many available, were chosen and applied to the two hierarchy levels. Two scenarios were simulated from Johnson Control Pte Ltd (JCS) product demand data. In scenario one, the Croston method of the module level was applied and after which converted to the component level, using it as a comparison against the direct forecast on the component level using the ARIMA method. In scenario two, a more regular intermittent demand, as compared to the first, was being created by the simulation of Croston’s data. The use of Minitab software helped to generate the mean square error (MSE) for the ARIMA model in which the lowest was selected to make a comparison against the Croston’s method. For the Croston’s model formulation, Microsoft excel’s solver was used to achieve the α coefficient by minimizing the MSE. Finally, to evaluate which of the two methodologies gave a better forecast of demand, the MSE of both methods were compared. The method with the lowest MSE is preferred as the difference between the actual observed values and forecast values is smaller and hence more accurate. From the results which were obtained from the first scenario, it was shown that ARIMA method was better and second scenario showed that Croston had more components with a lower MSE. However, a step further was taken and a third scenario which involved more data and with a more regular intermittent demand was carried out to strengthen the finding. It was shown that having sufficient data and regular intermittent demand will make Croston method even more accurate in forecasting.