煤矿瓦斯预警可视为是否安全的分类问题,数据呈现不平衡分布特点.为此,提出一种混合策略属性选择多决策树分类算法:算法融合代价敏感因子,结合C4.5和CART属性选择方法作为分裂指标,并采用了基于不同根节点信息的多决策树建树方法.首先采用11个非平衡数据集进行算法有效性验证,实验结果表明,该方法可以有效针对不平衡数据进行分类,保证高准确率的前提下,有效提高了少数类预测准确性;进而将该算法用于煤矿瓦斯数据预测,结果表明,所提出方法可以有效提高煤矿瓦斯数据的总体预测性能.%Coal mine gas early warning can be regarded as security classification problem,and the data show unbalanced distribution characteristics.Therefore,this paper presents a Cost-sensitive Hybrid Measure Attributes Selection Multi-Decision Tree (CHMDT) algorithm.It combines C4.5 and CART by hybrid measure as the attribute split selection method,which also considers cost-sensitive factor.The algorithm uses multi-decision tree building method based on different root node.The paper first uses 11 imbalanced data sets to illustrate the validity of the algorithm.Experimental results show that the proposed method can effectively deal with imbalanced datasets and improve the prediction accuracy of minority class under the high total accuracy performance.Moreover,the experimental results on coal mine gas early warning data show that the proposed algorithm can effectively improve the predicting performance of coal mine gas data.
展开▼
