新型混合重取样算法在岩爆预测中的应用  

A Novel Hybrid Re-Sampling Algorithm and Its Application in Predicting Rockburst

作　　者：谷琼[1,2] 蔡之华[2] 朱莉[2] 王贤明[3] 

机构地区：[1]襄樊学院数学与计算机科学学院,湖北襄樊441053 [2]中国地质大学计算机学院,湖北武汉430074 [3]温州大学瓯江学院信息系,浙江温州325035

出　　处：《地球科学（中国地质大学学报）》2010年第2期311-316,共6页Earth Science-Journal of China University of Geosciences

基　　金：国家高技术研究发展"863"计划(No.2009AA12Z117);襄樊学院规划项目(No.2009YA012)

摘　　要：针对岩爆现象发生的不均衡及发生机理受多因素影响的问题,在分析重取样技术的基础上,设计并实现了自适应选择近邻的混合重取样算法,并将其用于岩爆危险性预测.该方法结合过取样和欠取样方法的优势,改进了SMOTE过取样算法在产生合成样本过程中存在的盲目性及只能复制生成数值属性的问题,新算法能根据实例样本集内部分布的真实特性,自适应调整近邻选择策略,对不同属性的数据采取不同的复制方法生成新的少数类实例,控制和提高合成样本的质量;并通过对合成之后的数据集,用改进的邻域清理方法进行适当程度欠取样,去掉多数类中的冗余实例和边界上的噪音数据,减少其规模,在一定程度上达到相对均衡,从而,可有效地处理非均衡数据分类问题,提高分类器的性能.该算法在VCR采场岩爆实例上进行实验,预测的结果与实际情况完全一致,表明在工程实例岩爆危险性实例数据非均衡情况下实施混合重取样方案是可行的,预测准确率高,具有良好的工程应用前景.采用该方法可找到岩爆发生的主控因素,为深部开采工程的合理设计与安全施工提供科学依据.Because of poor understanding about the mechanism of rockbust and about the effect factors,the statistic data of large amounts of rockburst are typical imbalanced data sets （IDS）.On the basis of analyzing re-sampling technology,a novel hybrid re-sampling technique based on Automated Adaptive Selection of the Number of Nearest Neighbors （ADSNN-Hybrid RS） is proposed and applied to study the prediction of rockburst.This method takes advantage of both technology of improved Synthetic Minority Over-sampling Technique （SMOTE） method and Neighborhood Cleaning Rule （NCR） data cleaning method.In the procedure of over-sampling with the SMOTE method,blindfold new synthetic minority class examples by randomly interpolating pairs of closest neighbors were added into the minority class;and data sets with nominal features can not be dealt with.These two problems were solved by the automated adaptive selection of nearest neighbors and adjusting the neighbor selective strategy.As a consequence,the quality of the new samples can be well controlled.In the procedure of under-sampling,by using the improved under-sampling technique of neighborhood cleaning rule,borderline majority class examples and the noisy or redundant data were removed.The main motivation behind these methods is not only to balance the training data,but also to remove noisy examples lying on the wrong side of the decision border.The removal of noisy examples might aid in finding better-defined class clusters,therefore,allow the creation of simpler models with better generalization capabilities.In turn,it promises effective processing of IDS and a considerably enhanced classifier performance.The VCR rockburst data sets were employed as a sample IDS for classification and prediction.By adding extra artificial minority class samples as the expanded training set,experiment was conducted,which yields exactly consistent prediction results with the actual situation.The ADSNN-Hybrid RS and classification scheme we developed is feasible and reasonable for

关 键 词：岩爆 灾害 不均衡数据集 预测 合成少数类过取样 欠取样 

分 类 号：TU457[建筑科学—岩土工程]