基于支持向量机的概率积分法参数动态时序预报

根据支持向量机和时间序列分析原理，建立了开采地表移动的概率积分法参数动态时序预报模型。首先获得等时间间隔的地表移动参数数据，并进行时间序列的平稳性、零均值和正态性处理。通过合理确定模型嵌入维数和支持向量机参数建立支持向量机模型，采用平均绝对误差百分率和威尔莫特一致性指数两个指标对模型的精度和泛化性能进行评价。以淮北某矿实测地表移动参数为例，所预报的下沉系数、水平移动系数、主要影响角正切、开采影响传播角的绝对误差和相对误差较小，预报结果与实测结果吻合。

Time series forecasting of probability integration method parameters based on support vector machine

A dynamic forecasting model of probability integration method parameters was established according to theory of support vector machine(SVM) and time series analysis.Equal time interval surface movement parameters measured from observation station were obtained firstly and data were processed by stationary,zero mean and normality,embedding dimension and SVM parameters were chosen reasonably,MAPE and WIA were used as indicators to evaluate the accuracy and generalization performance.Data of an observation station in Huaibei coal mining area were taken as example,results show that maximum absolute error and maximum relative error of subsidence factor,displacement factor,main effect angle,and mining effect transference values are all small.All forecasting results are accurately and reliability which can meet the requirement of on site engineering.