论文部分内容阅读
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。
Most data processing problems in the chemical and chemical fields belong to the ill-posed problem in mathematics. However, traditional chemometrics algorithms such as linear and nonlinear regression and artificial neural networks neglect this feature. Solving it as a “well-posed problem” is a significant cause of “over-fitting” in data processing. In recent years, the newly proposed “support vector machine algorithm” is suitable for dealing with the problem of ill-posed, can limit over-fitting, and because of the use of kernel function algorithm, can effectively deal with nonlinear data sets, and the current chemical chemistry widely used artificial nerve Compared to the network, superiority is obvious, which has huge potential in chemical industry.