针对分段线性表示(即一阶线性函数表示)或分段常数表示(即零阶函数表示)在时间序列近似表示中拟合误差较大的问题,提出时间序列高阶函数分段表示方法.通过建立高阶函数候选集模型,利用拟合误差指标选取最优函数,为保证在分段点处连续,引入断点处约束条件;在分段点选取方式上,设定观测值变化阈值及分段区间阈值,保证压缩率的同时,保留重要点信息.实验结果表明,该算法相对于分段线性表示和分段常数表示,能更好地拟合原始序列.%For the problem of large fitting error in the time series approximation representation for the piecewise linear representation (the first order linear function representation) or the piecewise constant representation (the zero order function representation),the method of high order function segmentation representation was proposed.By setting up the candidate set model of higb order function,the optimal function was selected by fitting the error index.To ensure continuity at the segmentation point,breakpoints restrictions were introduced.In the selection of time series segmentation point,the observation change threshold and the interval threshold were set.By this way,the compression rate was guaranteed,important information was reserved.Comparing to the piecewise constant representation and the linear approximation,the simulation results show that the algorithm can better simulate the original sequence.
展开▼
