论文部分内容阅读
提出一种基于奇异值分解最小二乘法的自适应局部线性化预测方法.它要求数据矩阵的条件数不大于给定阈值,并据此自适应地确定当前相空间的维数,然后根据新的嵌入维数重构数据矩阵,进行模型的参数估计和计算当前预测值.实验结果说明所提方法精度高且稳健.特别是当嵌入维数接近最邻近向量的数目时,其性能显著优于普通局部线性化方法.
This paper proposes an adaptive local linearization prediction method based on the least squares method of singular value decomposition which requires that the number of conditions of the data matrix is not greater than a given threshold and accordingly determines the dimension of the current phase space adaptively, Embedding the dimension to reconstruct the data matrix, estimating the parameters of the model and calculating the current predicted value.The experimental results show that the proposed method has high precision and robustness, especially when the embedding dimension is close to the nearest neighbor vector, its performance is significantly better than normal Local linearization method.