Forecasting product design time based on Gaussian Margin Regression

摘要

In order to obtain a forecast model for product design time from a small data set, Gaussian Margin Regression (GMR) is presented on the basis of combining Gaussian Margin Machines and kernel based regression. Gaussian Margin Regression maintains a Gaussian distribution over weight vectors for kernel based regression. The algorithm is applied to seeking the least information distribution that will make actual value be included in the confidential interval with high probability, and embedded genetic algorithm is presented for choosing its relevant parameters. The results of application in injection mold designs reveal that the time forecast model based on GMR is of feasibility and validity.