A New Forecasting Method for Hard Disk Drive Manufacturing Throughput with a Hybrid Neural Network Model

摘要

In this research,a new prediction method for the production throughput forecasting for testing operation in hard disk drive manufacturing are presented.The method for forecasting is separated into three stages.The first stage is input variable selection which is selecting the key input variables for prediction model by using mutual information (MI) method.The second stage is developing the forecasting model with generalized regression neural network (GRNN).The last state is for increasing the forecasting accuracy by optimizing the smoothing parameter of GRNN.The comparison result with the current forecasting system in real factory has shown that the proposed method gives forecasting accuracy higher than current method.