Abstract: The traditional time series segmentation algorithm often ignore time characteristics of the time series, resulting in that fragmented results are not accurate enough. Therefore, a time series modeling representation algorithm based on hyperbolic tangent function constraint is proposed. The algorithm introduces the hyperbolic tangent function on the basis of the piecewise aggregation approximation and proposes the concept of the motion enhancement factor. Considering the impact of time, it completes the final time series segmentations by extracting the differences between the amounts of information in each subsequence. Experiments show that the algorithm has a small fitting error and can complete the macro similarity search by better utilization of the post-segmentation sequences and so on. Meanwhile, there is a characteristic that time series can grow dynamically. In the algorithm, both of the universality and accuracy are improved.

Key words: time series, piecewise linear representation, piecewise aggregation approximation, hyperbolic tangent function, mobile enhancement factor

摘要： 针对传统的时间序列分段算法往往忽略时间序列的时间特性，导致分段结果不够精确，对此，提出基于双曲正切函数约束的时间序列建模表示算法。该算法在分段聚合近似的基础上引入双曲正切函数并且提出了移动增强因子的概念，在考虑时间影响的基础上抽取出各个子序列所含信息量的差异完成最终的时间序列分段。实验表明该算法有较小的拟合误差，能够更好地利用分段后的序列，完成宏观的相似性查找等工作，并且满足时间序列动态增长的特点，算法的通用性、普适性、准确性均有所提高。

关键词: 时间序列, 分段线性表示, 分段聚合近似, 双曲正切函数, 移动增强因子