基于Stacking集成学习的采空区地面塌陷危险性预测

地面塌陷是采空区最主要的地质灾害，具有突发性、多发性、隐蔽性等特点，危害大且难以治理。针对煤矿采空区地面塌陷危险性预测因素复杂性和相关性的特点，结合陕西省神木市某煤矿采空区地面塌陷的实例，使用Stacking算法组合单一算法，综合分析地面塌陷影响因素，将采空区实测数据代入模型进行分析，对比单一算法的随机森林、提升树、逻辑斯谛回归预测模型，并将Stacking算法与各种改进的Stacking算法进行对比，对地面塌陷进行预测，建立了地面塌陷危险性评判模型及评判指标，为预测采空区引发地面塌陷灾害提供科学的依据。实验结果表明，运用Stacking集成学习方法的组合模型，预测精度有了明显提高。

Prediction of the risk of ground collapse in goaf based on Stacking integrated learning

Ground collapse was the most important geological hazard in the goaf.It had the characteristics of suddenness，frequent occurrence，concealment，etc.，which was harmful and difficult to manage.In view of the complexity and correlation characteristics of the predicting factors of the ground collapse risk in the coal mine goaf，it was combined with an example of ground collapse in a coal mine goaf in Shenmu City，Shaanxi Province，it was used the Stacking algorithm to combine a single algorithm to comprehensively analyze the influencing factors of ground collapse，substituted the measured data of the goaf into the model for analysis，and compare the random forest，lifting tree，and logistic regression prediction models of a single algorithm.And it was compared the Stacking algorithm with various improved Stacking algorithms to predict ground collapse，established a ground collapse hazard evaluation model and evaluation indicators to provide a scientific basis for predicting the ground collapse disasters caused by goafs.Experimental results showed that the prediction accuracy had been significantly improved by using the combined model of the Stacking integrated learning method.