基于随机森林模型的岩爆等级预测研究

以岩石单轴抗压强度、单轴抗拉强度、硐室最大切向应力、岩石压拉比、应力系数、弹性变形指数和完整性系数为岩爆评价指标, 建立了4种评价方案; 在引入随机森林算法的基础上, 建立了岩爆等级预测的随机森林模型, 并通过R语言编写代码对该模型进行了计算, 得出评价指标的重要性和预测结果; 将4种评价方案用随机森林法、线性回归法和支持向量机法分别进行预测并将结果进行对比分析。结果表明：随机森林法的岩爆预测准确率较高(达到97%), 适用于解决数据不完全的小样本问题; 岩石完整性系数重要度最大, 岩石单轴抗压强重要度最小。

Prediction of Rockburst Classification Based on Random Forest

Four evaluation programs were established with uniaxial compressive strength and uniaxial tensile strength of rocks, the maximum tangential stress, ratio of compressive strength to tensile strength of rock, stress coefficient, elastic strain index and integrity coefficient as evaluation indicators. A random forest model for prediction of rockburst classification was established using random forest calculation and was calculated with code by using R programming language, leading to the importance of evaluation indicators and prediction result. Then, four evaluation programs were predicated, respectively, with random forest, linear regression method and support vector machine, and the obtained results were compared and analyzed. It is found that the predication with random forest providing a high accuracy (up to 97%) can be used for those small samples with uncompleted data. And among all evaluation indicators, rock integrity coefficient shows to be the greatest important while uniaxial compressive strength being the least.