中文摘要：

介绍BP神经网络与多分类支持向量机等分类模型的基本原理,并基于这两种方法对水质识别与分类的准确度进行实例比较研究,随机抽取了南昌市内2010—2013年水域水质的300组数据为样本,选取了pH,氨氮,Cl-,SO2-3,总硬度,硝酸盐氮为评价的主要特征。通过把训练后的模型在测试集中进行的检验对得到的模型进行评估,表明了BP神经网络和多分类支持向量机均可以较好地解决水质识别与分类过程中存在的复杂性,多变量,非线性等问题,相比较而言多分类支持向量机有较强的鲁棒性,预测结果更为精确稳定,将其应用到水质评价中具有一定的可行性。

英文摘要：

This paper introduced the mechanism of BP Neural Network and Multi-class Support Vector Machine classification model,and the accuracy of water quality identification and classification was studied and compared based on these two methods.Many sets of data of water quality in Nanchang city during 2010-2013 were randomly selected as samples,and pH,Cl-,SO2-3,total hardness and nitrate nitrogen were selected as the main characteristics of the evaluation.The models was evaluated by testing the training models in the test sets.From the result we can find that the BP Neural Network and Multi-class Support Vector Machine can solve the problems of water quality identification and classification,such as the complexity,multi variable,nonlinear and so on.Compared with the BP neural network,the Multi-class Support Vector Machine has strong robustness and stability,and the prediction results are more accurate and stable,it is feasible to apply it to water quality assessment.