摘要
预测采煤工作面的瓦斯涌出量属于机器学习中的回归问题,主流方法包括CART和支持向量机等。CART决策树回归算法具有抽取规则简单、准确度高、可解释性强的优势,但是算法稳定性差,容易过拟合,同时每个叶节点的输出部分均为定值,难以动态的模拟真实数据的变化规律。支持向量机具有较好的鲁棒性,能够通过求解最小结构化风险来提高模型泛化能力,但是算法复杂度高,面对高维数据算法建模效率低下。结合支持向量机提出一种增强CART回归算法,最主要的改进是在每个叶节点的输出部分使用支持向量机建模。试验结果表明,与主流瓦斯涌出预测算法相比,增强CART回归算法能够有效提高采煤工作面瓦斯涌出量的预测精度。
From the view of data mining,predicting gas emission quantity from mining coal face is a kind of regression problem and major predicting algorithms include CART(classification and regression tree)and SVM(Support vector machine),and so on.CART has some inherent advantages,such as simple extraction rule,high accuracy and strong interpretability.However,CART is poor in stability and easily over-fitting.Moreover,the output of each leaf node is constant,which makes it difficult to dynamically simulate the variation law of real data.On the other side,SVM has good robustness and can improve the generalization ability by solving the minimum structural risk while its computation complexity is high and its efficiency is low when facing high-dimensional data.Combined with SVM,this article proposes an enhanced CART regression algorithm,of which the critical change exists in adopting SVM in each leaf node for model building.The experimental results show that the enhanced CART regression algorithm effectively improve the prediction accuracy of gas emission quantity from coal mining face.
作者
刘鹏
魏卉子
景江波
仰彦妍
LIU Peng;WEI Huizi;JING Jiangbo;YANG Yanyan(National Joint Engineering Laboratory of Internet Applied Technology of Mines,Xuzhou221008,China;Internet of Things Perception Mine Research Center,China University of Mining and Technology,Xuzhou221008,China;School of Information and Control Engineering,China University of Mining and Technology,Xuzhou221116,China)
出处
《煤炭科学技术》
CAS
CSCD
北大核心
2019年第11期116-122,共7页
Coal Science and Technology
基金
国家重点研发计划资助项目(2017YFC0804401,2017YFC0804409)
