改进极限学习机的不同类型滑坡位移预测

1.河南城建学院 测绘与城市空间信息学院,河南 平顶山 467036; 2.东华理工大学 江西省数字国土重点实验室,江西 南昌 330013

滑坡位移; 极限学习机; 预测

Various types of landslide displacement prediction based on the improved extreme learning machine
GAO Cai-yun1,2,GAO Ning1,2

(1.School of Geomatics and Urban Information,Henan University of Urban Construction,Pingdingshan 467036,China; 2.Jiangxi Province Key Lab for Digital Land,East China Institute of Technology,Nanchang 330013,China)

landslide displacement; extreme learning machine; forecasting

DOI: 10.13800/j.cnki.xakjdxxb.2018.0424

备注

针对经典智能算法用于滑坡位移预测时存在的网络结构参数选取复杂、易陷入局部极小等缺陷,提出了基于改进极限学习机ELM(Extreme Learning Machine)的滑坡位移预测模型。在滑坡变形位移状态辨识基础上,根据其位移变化特征,将滑坡位移曲线类型划分减速-匀速型、匀速-增速型、减速-匀速-增速型、复合型4类,将改进的ELM算法分别用于4种不同类型的滑坡位移预测。基于改进ELM算法构建滑坡位移预测模型时,采用二值区间搜索算法选定最佳隐含层神经元个数和激励函数,并融入数据滚动建模思想,以期提高网络泛化能力和预测精度。以链子崖、卧龙寺、古树屋、新滩滑坡体为例,对ELM预测的适用性进行讨论,实验结果表明,基于ELM构建不同类型滑坡位移预测模型时,具有较高的预测精度,且在网络学习速度等方面优势明显,适用于复杂状况下滑坡体的位移预测。

Considering complication to the parameter selection of conventional intelligent algorithm and being easy to fall into local minimum in prediction of landslide displacement,this paper proposes a new prediction model of landslide displacement based on improved Extreme Learning Machine.Characteristics of the landslide deformation accumulative displacement curve are studied and according to patterns,it is divided into four types:decelerating-uniform velocity,uniform velocity-accelerating,decelerating-uniform velocity-accelerating and compound type.The optimum number of neurons on hidden layer and excitation function of ELM are searched out according to the 2D range query-based and the technique of rolling modeling adopted in prediction in order to improve the prediction results.Taking landslides in Lianziya,Wolongsi,Gushuwu and Xintan examples,the applicability of improved ELM algorithm to landslide deformation displacement prediction is analyzed and compared.The experimental results show that ELM is valid and feasible in prediction of landslide under complicated conditions with higher precision.