基于粗糙集-神经网络的矿山地质环境影响评价模型及应用  

Study on the Model of Mines' Geological Environmental Impact Assessment Based on Rough Set and Artificial Neural Network and Its Application

作　　者：蒋复量[1,2] 周科平[1] 李书娜[1] 肖建清[1] 潘东[1] 李魁[1] 

机构地区：[1]中南大学资源与安全工程学院,长沙410083 [2]南华大学核资源与安全工程学院

出　　处：《中国安全科学学报》2009年第8期126-132,共7页China Safety Science Journal

基　　金："十一五"国家科技支撑计划项目(2006BAB02A02);湖南省安全生产科技发展计划项目(07-17;07-29;HN08-07);湖南省教育厅资助项目(07C652;08B068)

摘　　要：采用衡山白果地区石膏矿山的11个评价指标,综合运用粗糙集和神经网络理论,构建了基于粗糙集-神经网络(RS-ANN)的矿山地质环境影响评价模型,对RSES软件约简的数据和无约简的数据采用EasyNN-plus软件进行预测评价。神经网络模型的输入属性为8个,而粗糙集-神经网络模型的输入属性为6个,训练样本均为13个,预测样本均为4个,前者的平均预测精度为1.85%～24.86%,后者为1.23%～15.28%。研究发现,粗糙集在保留关键信息的前提下可有效地对数据表进行约简,约简后的神经网络预测结果与实际情况吻合,并比无约简时总体精度有较大幅度提高。Through referring to the 11 assessment indicators of gypsum mines in Baiguo region of Hengshan County, a model for mines＇ geological environmental impact assessment is set up based on rough set （RS） and artificial neural network （ANN）. Then, through adopting EasyNN-plus software, a prediction evaluation is made on the raw data and the data reduced by RSES software. The input attributes of the ANN model are 8, the RS-ANN model input attributes are 6, both training samples are 13, both forecast samples are 4, the former average prediction accuracy is 1.85% - 24.86%, the latter is 1.23% - 15.28%. This study shows that rough set is effective in the data table reduction while retaining key information ; the results predicted by RS - ANN model coincide with the actual situation, and the overall accuracy greatly rises.

关 键 词：矿山地质环境 评价模型 粗糙集 BP神经网络 评价指标 

分 类 号：X820.3[环境科学与工程—环境工程]