支持向量机及其他核函数算法在化学计量学中的应用  

Application of support vector machine and kernel function in chemometrics

作　　者：陈念贻 陆文聪 叶晨洲[2] 李国正[2] 

机构地区：[1]上海大学理学院化学系计算机化学研究室,上海200436 [2]上海交通大学图象及模式识别研究所,上海200030

出　　处：《计算机与应用化学》2002年第6期691-696,共6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)

摘　　要：化学、化工领域中多数数据处理问题属于数学中的“不适定问题”(ill-posed problem),而传统的化学计量学算法如线性和非线性回归,人工神经网络等忽略了这一特点,将其作为“适定问题”(well-posed problem)求解,是引发数据处理中“过拟合”问题的重要原因。近年来新提出的“支持向量机算法”适合于处理不适定问题,能限制过拟合,且因采用核函数算法,能有效处理非线性数据集,与当前化学化工中应用极广的人工神经网络相比,优越性明显,在化学化工中具有巨大的应用潜力。In the fields of chemistry and chemical engineering, most of the data mining problems are actually 'ill-posed problems' . But the traditional methods in chemometrics, such as linear or nonlinear regression and artificial neural networks, usually ignore the ill-posed charac-teristics and treat them as 'well-posed problems' . This ignorance usually induces significant overfitting problems. A newly proposed technique of data mining, called 'support vector machine' , is suitable for the data mining of ill-posed problems, without significant overfitting. Besides, since kernel function is used in this method, it is very suitable for the data mining of nonlinear data sets. Since this new method has significant advantages compared with ANN, which is now widely used in the fields of chemistry and chemical engineering, support vector machine exhibits great potentialities for many application topics in chemistry and chemical engineering.

关 键 词：核函数算法 化学计量学 应用 不适定问题 过拟合 支持向量机算法 化学 化工 数据处理 

分 类 号：O6-04[理学—化学] TQ015.9[化学工程]