| BP神经网络和SVM在矿山环境评价中的应用分析 |
| |
| 引用本文： | 李东,周可法,孙卫东,王金林,于浩,刘慧.BP神经网络和SVM在矿山环境评价中的应用分析[J].干旱区地理,2015,38(1):128-134. |
| |
| 作者姓名： | 李东 周可法 孙卫东 王金林 于浩 刘慧 |
| |
| 作者单位： | （1 中国科学院新疆生态与地理研究所，新疆矿产资源研究中心， 新疆 乌鲁木齐 830011； 2 中国科学院大学， 北京 100049；
3 新疆维吾尔自治区地质矿产勘查开发局信息中心， 新疆 乌鲁木齐 830000） |
| |
| 基金项目： | 国家自然科学基金(No.U1129302);新疆维吾尔自治区科技援疆计划项目(No.201090047) |
| |
| 摘 要： | 矿山环境的影响因素多样，定量评价过程易受人为因素干预。BP神经网络与SVM算法能够自动模拟各因子间的非线性关系。首次将其引入到矿山环境评价中，选取160个单元作为训练样本，以自然地理、基础地质、开发占地及地质环境等4个大类的14个变量指标为输入向量，以单元评价得分为输出向量，分别建立BP神经网络与SVM矿山环境评价模型。结果表明：两种模型均能满足矿山环境评价的精度要求；SVM模型收敛速度较BP神经网络快，MSE小于BP神经网络，更适合矿山环境评价工作；将定量模型应用于研究区，评价得分划分为4个级别，与定性评价结果一致，为矿山环境评价工作提供了新思路。
|
| 关 键 词： | 矿山环境评价 BP神经网络 支持向量机（SVM） GIS |
| 收稿时间： | 2014-06-18 |
Application of BP neural network and SVM in mine environmental assessment |
| |
| LI Dong;ZHOU Ke-fa;SUN Wei-dong;WANG Jin-lin;YU Hao;LIU Hui.Application of BP neural network and SVM in mine environmental assessment[J].Arid Land Geography,2015,38(1):128-134. |
| |
| Authors: | LI Dong;ZHOU Ke-fa;SUN Wei-dong;WANG Jin-lin;YU Hao;LIU Hui |
| |
| Institution: | （1 Xinjiang Institute of Ecology and Geography, Chinese Academy of Science, Xinjiang Research Center for Mineral Resources,Urumqi 830011, China ;2 University of Chinese Academy of Sciences, Beijing 100049, China; 3 Information Center of Xinjiang Bureau of Geology and Mineral Resources Exploration and Development, Urumqi 830000, Xinjiang, China） |
| |
| Abstract: | To gain environmental comprehensive evaluation objectively is one major subject of mine environment research，which contributes to sustainable development and utilization of local resources and the ecological environment restoration. There are many kinds of mine environmental impact factors and quantitative evaluation process is vulnerable to human factors intervention. BP neural network and SVM algorithm have the information processing ability and reasoning function to simulate the nonlinear relationship between each factor automatically. Based on remote sensing survey results of ore concentration in Qinghe Country，Xinjiang，China，the BP and SVM evaluation models，introduced to the mine environment evaluation for the first time，with 14 variables as input vector and unite score as output vector，have obtained good results. Both models select 160 units as the training sample that contains 4 large properties: natural geographical，basic geology，development covering and geological environment. The results show that both models can meet accuracy requirements of the mine environmental assessment （most of absolute error magnitude of the validation data is less than the grade e-1，77.5% and 95%，respectively）；MSE of SVM model （7.39 e-4） is smaller than that of BP neural network （1.5 e-3）；SVM model，whose convergence rate is faster than BP neural network，is more suitable for mine environmental evaluation. Environmental score is divided into 4 levels that is consistent with qualitative evaluation，so it indicates the model is worthy of promotion. |
| |
| Keywords: | mine environmental evaluation BP neural network Support Vector Machine GIS |
| 本文献已被 CNKI 万方数据 等数据库收录！ |
| 点击此处可从《干旱区地理》浏览原始摘要信息 |
| 点击此处可从《干旱区地理》下载免费的PDF全文 |
