Abstract: The state-of-art features extraction methods from time series are single-scale methods that result in the location of features imprecision and suppress the quality of discovered pattern. A novelty multi-scale features extraction methods from time series is proposed based on the principle of wavelet singularity detection. The time series are compressed into event sequence using singularity features and a dynamic time warping similarity measure of event sequenced is defined. The proposed algorithm is used to similarity pattern matching for event sequence. The experimental result shows that it has higher matching precision and lower computing cost.
【Key words】time series; similarity matching; singularity event; knowledge discovery

Key words: time series, similarity matching, singularity event, knowledge discovery

摘要： 现有的时间序列特征提取方法多为单尺度方法，导致特征点的时间定位不准确，从而影响模式发现的质量。该文基于小波奇异检测理论，提出了一种多尺度时间序列特征提取方法，利用奇异特征将时间序列压缩为事件序列表示，定义了事件序列动态时间弯曲相似度量，给出了基于事件序列相似模式匹配算法。实验表明，该方法具有较高的匹配精度和较低的计算代价。

关键词: 时间序列, 相似匹配, 奇异事件, 知识发现

CLC Number: 

• TP311