摘要
从石油录井色谱数据应用的实际需求出发,提出一种新的时间序列分段拟合算法。该算法通过一次扫描数据,根据中线距离阈值和非单调序列中极值点保持时间段阈值两个约束条件,选择反映序列趋势变化的关键点,然后线性拟合时间序列。实验结果表明该算法能够在保持原始序列主要形态的同时剔除噪音干扰,精确定位单调序列中的突变转折点,发现序列中的尖峰状态。
A new segmenting algorithm of time series was proposed to satisfy the requirement of processing chromatogram data of well-logging. The key points that denote the varying trend of the sequence were selected according to some thresholds through one scan of the data. These selected points were used to fit time series linearly. The experimental results show that this algorithm can eliminate the noises without loss of the rough shape, and find the inflexion and the peak subsequences accurately.
出处
《计算机应用》
CSCD
北大核心
2007年第7期1702-1704,共3页
journal of Computer Applications
关键词
时间序列
线性拟合
中线法
关键点
time series
linear fitting
midline length
key points