中文摘要：

针对污水处理过程生化需氧量（BOD）浓度难以实时监测的问题,建立了一种基于支持向量回归机（SVR）修正方法的案例推理（CBR）预测模型。该模型主要包括案例检索、案例重用、SVR修正、案例存储等4个部分,其中,SVR修正模型是利用历史数据构造修正案例库,并采用SVR训练而获得的,可以对传统CBR求解模型得到的BOD浓度建议值进行修正。实验表明本文模型的拟合误差优于支持向量机（SVM）、BP神经网络、RBF神经网络以及传统CBR方法,说明SVR修正方法的引入可以改善CBR的回归性能,提高CBR的学习能力。

英文摘要：

For the problem of monitoring biochemical oxygen demand （BOD） concentration in wastewater treatment process, a case-based reasoning （CBR） prediction model based on support vector regression machine （SVR） is established in this paper. This model is composed of a case retr ieval ,a case reuse,a SVR revision and a case retention. The SVR revision model is obtained using the SVR training to revise the BOD concentration suggested from the traditional CBR model. The experiment results indicate that the fitting error of this model is lower compared with the support vector machine （SVM）, the BP neural network, RBF neural network and the traditional CBR method. The application of SVR can effectively improve the regression performance and the learning ability for a traditional CBR model.