摘要
将支持向量回归(SVR)方法用于氧化铟薄膜的厚度控制。取已有的实验数据作为模式识别训练样本,以样品中氧化铟的重量百分含量、原料的粘度、添加剂的重量百分含量以及两个处理工艺条件提拉速度和提拉次数作为特征变量,得到了用于计算薄膜厚度的回归方程式。用“留一法”检验所得数学模型的预报能力,并将结果与传统的模式识别方法(Fisher法和KNN)进行了比较,结果表明:SVR的预报准确率比Fisher和KNN方法高。因此,SVR方法有望成为一种新的实验设计的手段。
The support vector regression (SVR) modei was used to control the thickness of In2O3 films, using the data from experimental results as training set, and some parameters such as the mass percentage of In2O3 and PVA, the viscosity of coating liquids, the drawing rate and drawing number as features. The regression equation for the computation of the thickness of films was obtained. The cross-validation experiment (leaving one method) was done in order to test the predictive ability of the SVR. The result of the experiment comparing the performance of the SVR with two other methods, Fisher discriminating vector method and KNN method indicated that the predictive ability of the SVR exceeded that of Fisher and KNN method.
出处
《计算机与应用化学》
CAS
CSCD
北大核心
2002年第6期733-736,共4页
Computers and Applied Chemistry
基金
国家自然科学基金委和美国福特公司联合资助(9716214)
