【摘要】 针对爆破振动效应评价中众多指标不确定性问题,基于支持向量机(SVM)原理和爆破振动对民房破坏的10个指标(爆破振动幅值、主频率持续时间、房屋高度等),利用Logistic回归分析理论,分析各指标与民房安全级别的关系,最终选取3个主要指标,建立露天采矿爆破振动对民房破坏的SVM预测模型。以32组实测数据作为学习样本对该模型进行训练。采用回代估计法对该模型进行回检,误判率为0。利用对小样本数据有较好预测性能的该模型进行另外12组检验样本的预测。结果表明:经回归分析,指标个数缩减,相关性降低,SVM运算效率提高,回估误判率为16.667%。更多还原

【Abstract】 Seeing that commonly used methods for evaluating effects of blasting vibration were faced with so main indexes uncertainty problems,a model,based on SVM,was built for predicting residential houses’ damage from blasting vibration of open pit mining. Before ten factors and the houses’ safety levels had been analyzed using Logistic regressive analysis theory,and three main indexes had been selected. The re-substitution method was used to verify the stability of the model( false rate was 0),and the model was used to discriminate twelve new samples. The results show that the number of indexes was rendered smaller and correlation among indexes was rendered lower by logistic analysis,the efficiency is improved when the algorithm of SVM is operating,the re-substitution false rate is 16. 667%.更多还原