改进的主成分分析法在多响应优化中的应用

Application of Improved Principal Component Analysis Method to the Multi-Response Optimization

提出了一种改进的主成分分析法,在解决多响应优化问题时考虑到模型的预测能力。主成分分析法是一种常用的多响应优化方法,为了在应用主成分分析法的过程中考虑到模型的预测能力,文章将回归方程拟合度R2系数结合到主成分分析中。与主成分分析法相比,该方法不仅能利用主成分得分将多个响应转化为单一响应,还能考虑到不同响应的预测能力。实例表明,用该方法得到的结果可体现出模型预测能力的影响,并且预测能力强的响应得到较大的改进。

An improved principal component analysis method, which takes into consideration the differ ence in the predictive ability among the responses is proposed. Principal component analysis method is a popular method for multiresponse optimization. To consider the predictive ability, the proposed method applies R2 coefficient which represents the degree of fit of the regression model to principal component a nalysis. Compared with the existing principal component analysis method, the proposed method can transform multiresponse into single response, and consider the predictive ability. Case study shows that the results obtained through this method can reflect the impact of the predictive ability, the response with stronredictive ability is reatlv improved.