Abstract: Coal and gas outburst is one of the principal dynamic disasters in coal mine underground mining. Aimed at solving the prediction problem of coal and gas outburst risk level, improve the outburst prediction accuracy, this paper selects 9 factors as evaluation index which affects the coal and gas outburst classification prediction, including maximum principal stress, gas pressure, gas content, roof lithology, distance from the fracture of roof, coal seam thickness, mining depth, the absolute gas emission and relative gas emission, and do the factor analysis to the most related evaluation index, extract the public factor, with random forest algorithm training, establish the coal and gas outburst prediction model based on factor analysis and random forest. Based on 19 groups of coal and gas outburst data measured by coal mine as the training sample data sets for model training, 5 groups of data as test data for model testing, do the prediction of coal and gas outburst, meanwhile compared with other prediction models, the test results verify the random forest algorithm in coal and gas outburst prediction has higher accuracy.

Key words: random forest, factor analysis, coal and gas outburst, strength prediction

摘要： 煤与瓦斯突出是煤矿开采过程中主要的动力灾害之一，针对煤与瓦斯突出等级预测问题，提高突出预测的准确率，选取最大主应力、瓦斯压力、瓦斯含量、顶板岩性、距断裂距离、煤层厚度、开采垂深、绝对瓦斯涌出量和相对瓦斯涌出量9个影响因素作为煤与瓦斯突出等级预测的评价指标，同时对相关程度较高的评价指标进行因子分析，提取公共因子，用随机森林算法进行训练预测，建立了基于因子分析的煤与瓦斯突出预测的随机森林模型。通过煤矿实测19组煤与瓦斯突出的数据作为训练样本数据集进行模型的训练，5组数据作为该预测模型的测试数据，进行煤与瓦斯突出预测，同时通过其他预测模型预测结果的对比，验证了随机森林算法在煤与瓦斯突出预测中具有较高的准确度。

关键词: 随机森林, 因子分析, 煤与瓦斯突出, 强度预测