论文研究-时间序列度量的斜率偏离距离方法研究.pdf

所需积分/C币:10 2019-09-10 17:38:53 387KB .PDF
32
收藏 收藏
举报

针对数据挖掘领域中时间序列的相似性度量问题,提出一种斜率复合偏离距离方法。以大量噪声的高维多元时间序列数据为目标,提出了一种基于斜率偏离度的时间序列相似性度量方法。该方法主要是在分段线性的基础上,基于角度和斜率进行偏离度计算,解决普通斜率距离度量的局限性,物理意义更为明确,实际度量更为准确。证明了斜率复合偏离的完备性和连续性,最后用仿真算例对算法的有效性进行了验证。
62009,45(22) Computer Engineering and Applications计算机工程与应用 6000 备性和连续性都得到了证明,从仿真算例中可以看出,斜率复 4000 合偏离可以把曲线之间的趋势有效表现出来 2000 05001000150020002500s 参考文献: [1 Chen M S, Han J, Yu P SData mining: An overview from a database perspective[J]. IEEE Transacions of Knowledge and Data Engineering,1996,8(6):866-883 [2 Sidiropoulos N D, Bros R Mathematical programming algorithms 50020002500t/s for regression-based non-linear filtering in n-dimensional real space[J].IEEE Transaction on Signal Processing, 1999, 47(3): 771-78 400 3 Agrawal R, Lin K I, Sawhney H S,et al. Fast similarity search in 200 1000150020002500t/s the presence of noise, scaling, and translation in time-series 图4三种股票指数序列曲线图 database[ C]//Proceedings of the 21st International Conference on Very Large Data Bases. San Francisco CA USA: Morgan Kaufmann 表1不同分段时的斜率复合偏离度 Publishers Inc, 1995: 490-501 段数Dms(S1,S2)Dats(S1,S3)Dans(S2,S3) [4 Berndt D J, Clifford J Using dynamic time warping to find patterns 10.683 in time series[ C]/Proceedings of the KDD Workshop, Seattle, WA 0~2800 12.353 4.2090 15.973 1994:359-370 19.352 5.1532 23.253 5]王达,荣冈时间序列的模式距离J浙江大学学报:工学版,2004, 26.353 5.2350 33.253 段数Dms(S1,S2)Dmsk(S1,S3)Das(S2,S3) 38(7):395-398 6.8662 9.153 6 Faloutsos C, Ranganathan M, Manolopoulos Y Fast subsequence 10.3530 3.321 13.353 matching in time series databases[ C]/Proc of the ACM SIGMOD 13.2530 3.653 17.223 International Conference on Management of Data. Mineapolis: ACM 90 19.0350 3.987 23.354 Press,1994:419-429 离递增幅度大大增加,斜率距离的优势不再明显。 [7 Bayer K, Coldstein J, Ranlakrishnan R, et al. When is nearest neigh- 从表1的数据可以看出,使用斜率复合偏离度比较后,随 bors meaningful? [C]/The 7th Int'l Conf on Database Theory Jerusalem, Israel, 1999 着分段数量的增加,相似的两条曲线S1与S3之间的斜率偏离 [8] Keogh E J, Pazzani M J An indexing scheme for fast similarity 度趋向一个比较稳定的值,不相似的曲线之间的距离随着分段 large time series database[ C//Proceedings of Interna 数量的增加逐步变大,结论合理,符合人对曲线关系的认识。 tional Confcrcncc on Data Mining. Clcvcland: IEEE Computcr Soci ety,1999:56-67 6结束语 9]张建业,潘泉,张鹏基于斜率表示的时间序列相似性度量方法J 提出了基于斜率偏离度量的时间序列距离度量方法,其完 模式识别与人工智能,2007,20(2):271-274 (上接3页) ing. San Francisco: Morgan Kaufmann. 1996: 71-77 着测试运行时间过长,分类准确率不高的不足。提出了基于[7] Romdhani s,lorP, Scholkopf B,etal. Computationally efficient LDE和简化SⅤM的文档分类算法,实验结果表明该算法具有 face detection[C]//Proceedings of the Eighth IEEE International 很好的分类性能和很快的测试运行速度。 Conference on Computer Vision. Piscataway: IEEE Press, 2001 参考文献: [8] Salton G, Wong A, Yang C SA vector space model for automatic [1] Sebastiani F Machine learning in automated text categorization [JI indexing]. Communications of the ACM, 1975, 18(11): 613-620 ACM Computing Surveys, 2002, 34(1): 1-47 9 Berry M W, Drmac Z, Jessup E R Matrices, vector spaces, and in- 2]王自强,钱旭,孔敏流形学习算法综述[计算机工程与应用 formation retrieval[J] .SIAM Review, 1999, 41(2): 335-362 2008,44(35):9-12 [10 Duda R O, Hart P E, Stork D G Pattern classification[M].2nd ed [3] Chen H T, Chang H W, Liu T L Local discriminant embedding Hoboken Wiley-Interscience, 2000 and its variants[C]//Proceedings of 2005 IEEE Computer Society [11] Hsu C W, Lin C J.A comparison on methods for multi-class Conference on Computer Vision and Pattern Recognition. Piscat support vector machines(JIEEE Transactions on Neural Networks away: IEEE Press, 2005: 846-853 2002,13(2):415-425 [4] Vapnik V N The nature of statistical learning theory[M).New York: [12] Lewis D D Reuters-21578 text categorization collection [EB/OL Springer, 1995 11999-01-10].http://kdd.ics.uci.edu/databases/reuters21578/reuters- [5] Scholkopf B, Mika S, Burges C J C, et al. Input space versus fea 21578.html ture space in kernel-based methods[J].IEEE Transactions on Neu- [13 Craven M, DiPasquo D, Freitag D, et al. Learning to extract sym- ral Networks,1999,10(5):1000-1017 bolic knowledge from the World Wide Web[c]//proceedings 16 Burges C J CSimplified support vector decision rules[Cy/proceed- the Fifteenth National Conference on Artificial Intelli ings of the Thirteenth International Conference on Machine Learn bridge: AAAI Press/The MIT Press, 1998: 509-516

...展开详情
试读 3P 论文研究-时间序列度量的斜率偏离距离方法研究.pdf
立即下载
限时抽奖 低至0.43元/次
身份认证后 购VIP低至7折
一个资源只可评论一次,评论内容不能少于5个字
您会向同学/朋友/同事推荐我们的CSDN下载吗?
谢谢参与!您的真实评价是我们改进的动力~
  • 至尊王者

关注 私信
上传资源赚钱or赚积分
最新推荐
论文研究-时间序列度量的斜率偏离距离方法研究.pdf 10积分/C币 立即下载
1/3
论文研究-时间序列度量的斜率偏离距离方法研究.pdf第1页

试读结束, 可继续阅读

10积分/C币 立即下载