摘要
鉴于生物发酵过程的高度非线性,且样本采集困难,数据总量较少等,采用支持向量机(SVM)方法,为柠檬酸发酵过程建模,得到最终酸度与相关因素间的定量关系。通过优化建模参数,所建SVM模型具有较高的拟合能力,且预测误差小,稳健性好。实例表明,与人工神经元网络等方法相比较,SVM方法更为优越。
Support Vector Machines (SVM) was applied to set up the model of the relationship between the final concentration of citric acid and the relative factors considering the difficulties of collecting experimental samples and the lack of the amount of data and the problem of severe non-linearity in fermentation process. The model optimized the parameters and was compared with that made by Artificial Neural Networks (ANN). The experimental results showed that the model based on SVM has high fitting abilities as well as ANN and has less prediction errors and less standard deviation of prediction errors than ANN.
出处
《化学反应工程与工艺》
CAS
CSCD
北大核心
2004年第1期59-63,共5页
Chemical Reaction Engineering and Technology
基金
国家自然科学基金资助项目(编号:20076041)。
