Abstract: To apply support vector machine(SVM) for evaluation of eco-environment,it is essential to give priority to designing the classifier and picking kernel functions,their parameters and penalty factors.Described here are the ways of choosing decision-tree as SVM multi-class classification method and radical base kernel functions to build a decision-tree-based SVM evaluation model for eco-environment level.Self-adaptive optimization of penalty factors and kernel function parameters is realized.The evaluation model is tested for eco-environment evaluation of some major cities and provinces of China.The test demonstrates that SVM is a good method to evaluate eco-environmental quality.