摘要
多变量时间序列(MTS)在金融、多媒体、医学等领域的应用是非常普遍的.与其它多变量时间序列样本显著不同的样本,我们称之为异常样本.本文提出一种基于局部稀疏系数的多变量时间序列异常样本的识别算法,使用扩展的 Frobenius 范数来计算2个 MTS 样本之间相似性.使用两阶段顺序查询来进行 k-近邻查找,将不可能成为候选异常样本的 MTS 样本剪去.在2个实际数据集上进行实验.实验结果验证算法的有效性.
Multivariate time series （MTS） datasets are commonly used in the fields of finance, multimedia and medicine. MTS samples, namely outlier samples, are significantly different from the other MTS samples. In this paper, a method for detecting outlier samples in the MTS dataset based on local sparsity coefficient is proposed. An extended Frobenius norm is used to compare the similarity between two MTS samples, and k -nearest neighbor （ k -NN） searches are performed by using two-phase sequential scan. MTS samples that are not possible outlier candidates are pruned, which reduces the number of computations and comparisons. Experiments are carried out on two real-world datasets, stock market dataset and BCI （ Brain Computer Interface ） dataset. The experimental results show the effectiveness of the proposed method.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
2007年第4期463-468,共6页
Pattern Recognition and Artificial Intelligence
基金
国家自然科学基金(No.60173058)
