基于二项logistic回归模型与CART树的煤层底板突水预测.pdfVIP

第37 卷 第 1 期 煤田地质与勘探 Vol. 37 No.1 2009 年2 月 COAL GEOLOGY EXPLORATION Feb. 2009 文章编号: 1001-1986(2009)01-0056-06 基于二项 logistic 回归模型与CART 树的 煤层底板突水预测 刘再斌，靳德武，刘其声 (煤炭科学研究总院西安研究院，陕西 西安710054) : 为定量评价煤层底板突水信息对突水过程的影响程度，获得煤层底板突水规则，采用二项 logistic 回归与CART 树相结合的方法进行煤层底板突水预测。在煤层底板突水信息分析的基础上， 建立了包含全因素的煤层底板突水预测概率模型，基于向后逐步回归分析方法获得了包含 项主 6 要突水信息的精简煤层底板突水预测概率模型。通过 CART 树算法获得了煤层底板突水规则，分 类测试结果表明，所获得的突水规则分类准确率达到91.67% 。 ：二项logisitic 回归；突水预测；突水信息；CART 树 ： ： TD741 A Prediction of water inrush through coal floor based on binary logistic regression model and CART LIU Zaibin, JIN Dewu, LIU Qisheng (Xian Branch , China Coal Reaserach Institute, Xian 710054, China) Abstract: To quantitatively evaluate the information of water inrush through coal floor, and get rules of water inrush, binary logistic regression and classification and regression tree were used to predict water inrush through coal floor. Based on the analysis of water inrush information, a probabilistic predicting model which contained all selected fac- tors was established, a simplified model which contained six main factors was established based on backward stepwise regression method. Rules of coal floor water bursting were found by classification and regression tree method, the re- sult of classification test showed that the classification accuracy of water inrush achieves up to 91.67%. Key words: binary logistic regression; water inrush prediction; water inrush information; classification and regression tree logistic