APPLICATION OF OPTIMAL DIMENSION MODELING FOR BP NETWORK IN PREDICTION OF TIME SERIES OF DEFORMATION
Peng Xiangguo1，2）;Wang Jian1，2） ;and Sun Lihong1，2）
1）Jiangsu Engineering Key Laboratory of Resource and Environment Information，Xuzhou 2210082）School of Environment and Spatial Informatics，China University of Mining and Technology，Xuzhou 221008
Abstract It is difficult for BP network to determine the number of input layer’s nodes as dealing with the deformation data of single time series.In order to solve this problem，the GP algorithm is used to calculate the correlation dimension.Taking the embedding dimension m as the best model dimension and reconstructing an equivalent phasespace according to the theory of reconstruction so as to provide a reasonable input for BP network.Considering the influence of dynamical factor of deformation as well as the difficulty of calculating number of input cell，the model can be applied to simulate and forecast the deformation.Two actual examples show that this model is reasonable and precise.
Peng Xiangguo,,Wang Jian et al. APPLICATION OF OPTIMAL DIMENSION MODELING FOR BP NETWORK IN PREDICTION OF TIME SERIES OF DEFORMATION [J]. , 2009, 29(6): 113-117.
Peng Xiangguo,,Wang Jian et al. APPLICATION OF OPTIMAL DIMENSION MODELING FOR BP NETWORK IN PREDICTION OF TIME SERIES OF DEFORMATION [J]. jgg, 2009, 29(6): 113-117.
