Defective Recognition with MTD-based Non-Equigap Grey Models.

The TFT-LCD (thin-film transistor liquid crystal display) industry in Taiwan has been developed for decades. To provide better service for customers, most Taiwan companies in the supply chain have set up their warehouse in mainland China for vendor managed inventory (VMI). However, over the last decade, owing to the industry issues of product diversity strategies and short product lifecycles, it has become difficult for suppliers to control well the stock level in VMI, especially the operation of defective product returns. In order to keep high customer satisfaction, it is considered an option by forecasting possible future defectives for preparing the returning products to satisfy customers. The non-equigap grey model (NGM) used to be applied to such short-term time series data and has shown to be an effective tool. However, NGM predictions still can be improved by making more appropriate background values by determining their parameter a values. Accordingly, this paper employs the mega-trend-diffusion (MTD) technique to estimate better a values, and the model is thus called MTD-NGM(1,1). In the experiments, two data sets taken from a supplier are examined for effectiveness validation. The experimental results indicate that MTD-NGM(1,1) can generally produce better predictions than NGM(1,1). [ABSTRACT FROM AUTHOR]