Abstract: Time series is an important high dimension data type composed of a series of data observed according to time sequence.Usually time series data is massive.Sometimes，it is inefficient or impractical to mine the original time-series data directly.So time series should be represented on a higher level.Reference to the idea of piecewise linear of time series，we put forward a method of adaptive piecewise linear representation of time series based on error restricted.This method can give the number of fitting lines automatically according to a given error e on the special points of time series，This method not only compresses data and wipes of noise，but also finds the various pattern features of time series.According to experiments our algorithm leads to less fitting error and better performance than other algorithms.

Key words: data mining, time series, adaptive, piecewise linear representation

摘要： 实际过程中采集到的时间序列数据通常是海量数据，在原时间序列数据上直接进行数据挖掘的效率通常是低下的，有时甚至不可行，因此就须将时间序列在更高的层次上进行表示。借鉴时间序列线性分段的基本思想，提出了一种自适应误差约束的分段线性表示方法，该方法在查找出时间序列特殊点的基础上，通过给定误差e进行调节，可以自动地产生拟合线段的数目。不仅可以压缩数据，去除噪声，还能得到时间序列的模式变化特征。与一般的分段线性表示相比，文中方法的拟合误差更小，适应能力更强。

关键词: 数据挖掘, 时间序列, 自适应, 分段线性表示