Using GA- and MTD-based Non-Equigap Grey Models for Predicting Defective Item Quantities of Vendor Inventory

Abstract

The TFT-LCD (thin film transistor liquid crystal display) industry is the mainstream of flat panel display development in Taiwan However in order to provide better services to customers most Taiwanese companies have established their warehouse in China for vendor-managed-inventory (VMI) Due to industry characteristics product diversification and short product life cycle it is difficult for suppliers to control VMI's inventory level especially for the control of defective product returns The prediction on future inventory defective product can be consider in order to avoid shortage inventory and maintain high customer satisfaction The non-equigap grey model (NGM) is applied to short-term time series data and has proven to be an effective tool By determining the value of the parameter (? values) to obtain a more appropriate background value the NGM prediction can be effectively improved This paper uses the mega-trend-diffusion (MTD) technique to learn the initial value of ? and then uses genetic algorithm (GA) to adjust these values to the optimal This NGM with the best alpha value is therefore called GA-MNGM (1 1) In the experiment two real-world case data sets collected by TFT-LCD suppliers were used for validity verification The experimental results show that the overall result of GA-MNGM(1 1) is better than MNGM(1 1) and its extended version

Date of Award	2019
Original language	English
Supervisor	Der-Chiang Li (Supervisor)