Development of scrap cost forecasting model: A case study in hard disk drive company

Abstract

The objective of this research is to develop the optimized scrap cost forecasting model in the Hard Disk Drivemanufacturing with more accuracy. Due to the current forecasting, calculated from the Key Performance Index(KPI) of all assembly processes, has error about 30% which means the production planning is imprecise and cannot specify the cause of scrap due to some KPIs are irrelevant with scrap cost. This is the reason why thisresearch aims to improve accuracy more than 10%. At first, the correlation analysis is used to identify the relationshipbetween KPI and scrap cost. The time-series analysis of KPI is taken and forecasted value of KPI is used asindependent variables in the scrap cost forecasting model formulation. Two techniques are applied to formulate forecasting models: regression analysis and artificial neural networks. The results from scrap cost forecasting ofproduct A, B and C have showed that the artificial neural network model can forecast with more accurate thanregression model. The forecasting errors of artificial neural network models are 11.48%, 11.43% and 18.86% forproduct A, B and C respectively.

Keywords : Forecasting, Correlation analysis, Regression analysis, Artificial neural networks