生态环境质量诊断和安全评价是保障矿区经济高质量发展的重要组成部分。针对当前矿区生态安全评价方法精度不高的问题，从生态环境质量、污染物排放、生态保护、环境承载力和系统协调5个方面构建评价指标体系。通过核函数选择，构造广义最优分类超平面，将小波理论和支持向量机方法有机结合，建立联合评价模型，并运用于G稀土矿区进行实例验证。结果表明：该稀土矿区Ⅰ级风险区域（即环境差区）有10个，Ⅱ级风险区域（即环境较差区）为8个。与GIS识别结果比较，模型的误差率为5%，说明模型具有较好的预警精度。模型整体预测性能较优，在非线性时间序列领域具有很好的表现和应用前景。

In the primary stage of mining area development，economic growth should be realized by relying on resource endowment of mining area.The high consumption of resources and energy and the high input of production factors give rise to the huge cost of environmental damage in the mining area，the pollution problems are evolving from simple one-way to complex ones.The system of compensation for ecological and environ-mental damages is conducive to solving the law enforcement dilemma of “corporate pollutes， residents victims， government pays”，and helping China’s ecological and environmental protection to the depth of development.However，ecological safety evaluation can provide a basis for the diagnosis of ecological environment quality and compensation for damages in mining areas，and they are also an important part of high-quality development of mining area economy.Aiming at the low accuracy of themining ecological security evaluation，the index system for mine ecological security evaluation was designed，including mining environment quality，pollutant discharge or emission，ecological protection，carrying capacity and the coordination of ecological system.Furthermore，20 secondary indexes were set，such as vegetation coverage rate，pollution source diffusion rate，land reclamation rate and so on.The quasi-optimal classification hyper-plane was constructed based on selected kernel function to combine Wavelet Theory with Support Vector Machines，and an integer hybrid model was established.The evaluation model was applied to the ‘G’ rare earth mining area，and the results show that there are 10 samples with grade Ⅰ risk level，that is，their environment are poor，and there are 8 samples with grade Ⅱ risk level，that is，the environment is poorer.Compared with results by GIS identifying，the error rate of the model is 5%，which has better warning accuracy.Further，compared with BP neural network，ARIMA and other methods commonly used at present，the proposed model has well-forecasting performance，can represent the nonlinear time series well.So，it has a good application prospect in the field of nonlinear time series.

http://www.goldsci.ac.cn/article/2020/1005-2518/1005-2518-2020-28-6-902.shtml