Method for Product Design Time Forecasting Based on Support Vector Regression with Probabilistic Constraints

There exist problems of small samples and heteroscedastic noise in design time forecast. To solve them, support vector regression with probabilistic constraints (PC-SVR) is proposed in this article. The mean and variance functions are simultaneously constructed based on a heteroscedastic regression model. Probabilistic constraints are designed to make sure that for every sample, the forecast value is in a neighborhood of the target value with high probability. The optimization objective is formatted in the form of par-v-SVR. Prior knowledge about maximum completion time can be embedded in probabilistic constraints, and provides the size of the neighborhood of the target value. The results of application in injection mold design have confirmed the feasibility and validity of PC-SVR.