摘要
传统支持向量机(SVM)评价模型中网格搜索法对参数的选择受到主观因素的影响,选用粒子群算法对SVM模型进行优化,并应用改进的SVM模型(PSO-SVM)对长吉图经济区135个矿山进行地质环境质量评价。PSO-SVM模型的评价结果与综合评价结果的相同率(注:该相同率是指两个评价结果相同的个数占所有评价样本数的百分比)达到95.56%,与SVM模型评价结果的相同率达到91.11%。结合研究区实际情况并分析三种评价结果得出,PSO-SVM模型的评价结果更符合实际情况。改进的支持向量机方法能够避免人为因素影响,提高矿山地质环境评价水平,在评价中具有可行性和有效性。基于改进的支持向量机方法评价结果表明,研究区矿山地质环境受矿山开采等人为活动的影响,54.1%的矿山遭受严重破坏(III级),25.9%为中度破坏(II级)。评价结果可为研究区矿山环境恢复治理提供决策支持。
The grid search method, as on kind of traditional evaluation model of SVM, can be influenced by subjective factors on selecting parameters. So in this study, particle swarm algorithm was chosen to optimize the SVM model and optimized SVM model （PSO-SVM） was used to evaluate geological environment of 135 mines in the Chang-Ji-Tu economic zone. It showed that the evaluation results were almost consistent with comprehensive evaluation results, the similiarity reached to 95.56%. Compared to traditional SVM, itreached to 91.11%. Based on the actual situation and comprehensive analysis of three kinds of evaluation results, we found out that PSO-SVM evaluation results were more in line with the actual situation. Optimized SVM model could effectively avoid the influence of human factors and improve the level of mine geological environment assessment, it was feasible and effective in the evaluation. Evaluation results based on the optimized SVM model showed that mine geological environment was affected by human activities in the study area such as mining, 54.1% of the mine suffered serious damage （III grade）, 25.9% suffered moderate damage （II level）. The evaluation results can provide decisions for environmental restoration in the study area.
出处
《吉林大学学报（地球科学版）》
EI
CAS
CSCD
北大核心
2016年第5期1511-1519,共9页
Journal of Jilin University：Earth Science Edition
基金
中国地质调查局项目(1212011140027
12120114027401)
吉林大学研究生创新基金资助项目(2015083)~~
