改进的决策树支持向量机地下水水质评价

基于结构风险最小原理的支持向量机(SVM)具有较强的学习泛化能力和良好的分类性能，能用来解决少样本学习的二类模式识别问题。针对具备多级类别的地下水水质评价问题，可以采用决策树SVM分类方法,通过对多类别水质标准的重新组合以构建类似于决策树的多个子分类器来实现。但基于决策树SVM分类过程中常常会出现由于正负类训练样本数据不均一导致的局部识别误差。基于二叉树原理提出了一种改进决策树SVM模型，通过加密数据插值和二叉分类有效避免正负类训练样本数据不均一的问题，针对地下水水质评价特点,增加了第5个子分类器以精确识别Ⅱ类水质和Ⅲ类水质。实验结果表明，改进的决策树SVM分类模型评价结果稳定。

Groundwater quality evaluation based on optimized model of decision-tree-based support vector machine

Support Vector Machine (SVM) based on the minimum of structured risk is characterized with strong ability to learn and predict and favorable classification performance, which makes it able to solve the two types of pattern recognition of fewer sample learning. In order to evaluate the groundwater quality which has five classes with SVM, the decision-tree-based way of rebuilding the classes like decision tree to create more sub two-class SVM would be used. But as a solution of classifying more classes, some defaults exist in decision-tree-based support vector machine (DTBSVM) including the local error produced by different sample mount between two classes. The authors brought forward an optimized DTBSVM model based on the principle of two cross tree to realize the evaluation for groundwater quality. The experimental results show that the optimized DTBSVM model is a good way to evaluate the groundwater quality.