基于高斯间距核回归的产品设计时间预测

为克服产品设计时间预测中的小样本和异方差噪音问题,建立一种基于高斯间距核回归(Gaussian margin kernel regression,GMKR)预测模型。首先,假定核函数回归模型的权重向量服从高斯分布,利用相对熵与输出概率密度的自然对数和设计优化目标,构建GMKR模型;然后,假设高斯分布的协方差阵为对角矩阵以简化GMKR模型,并利用粒子群算法求解相应优化问题。最后,以注塑模具设计的实例进行分析,结果表明基于GMKR的时间预测模型可行有效。

Product Design Time Forecast by Using Gaussian Margin Kernel Regression

There exist problems of small samples and heteroscedastic noise in design time forecast. To solve them, Gaussian margin kernel regression （GMKR） is proposed. First, the Gaussian distribution over weight vectors for the kernel - based regression is assumed for GMKR, and the optimization objective function of GMKR is designed by considering both the relative entropy and the sum of the natural log of the output probability densities. Then, the optimization problem of GMKR is simplified by assuming the covariance matrix of the Gaussian distribution to be a diagonal matrix, and its relevant optimization problem is solved based on particle swarm optimization algorithm. Finally, the effectiveness of GMKR is verified by our experiment results on the time forecast of plastic injection mold design.