基于融合歐氏距離與Kendall Tau距離度量的譜聚類算法(英文)
發(fā)布時間:2018-12-12 21:42
【摘要】:大多數(shù)現(xiàn)存的譜聚類方法均使用傳統(tǒng)距離度量計算樣本之間的相似性,這樣僅僅考慮了兩兩樣本之間的相似性而忽略了周圍的近鄰信息,更沒有顧及數(shù)據(jù)的全局性分布結構.因此,本文提出一種新的融合歐氏距離和Kendall Tau距離的譜聚類方法.該方法通過融合兩兩樣本之間的直接距離以及其周圍的近鄰信息,充分利用了不同的相似性度量可以從不同角度抓取數(shù)據(jù)之間結構信息的優(yōu)勢,更加全面地反映數(shù)據(jù)的底層結構信息.通過與傳統(tǒng)聚類算法在UCI標準數(shù)據(jù)集上的實驗結果作比較,驗證了本文的方法可以顯著提高聚類效果.
[Abstract]:Most of the existing spectral clustering methods use the traditional distance measure to calculate the similarity between samples, which only considers the similarity between pairwise samples and neglects the neighboring information, and does not take into account the global distribution structure of the data. Therefore, a new spectral clustering method combining Euclidean distance and Kendall Tau distance is proposed. By combining the direct distance between two samples and the adjacent information around it, the method makes full use of the advantages of different similarity measures to capture the structural information between data from different angles. A more comprehensive reflection of the underlying structure of the data information. By comparing with the experimental results of traditional clustering algorithm on UCI standard data set, it is verified that the proposed method can significantly improve the clustering effect.
【作者單位】: 南京航空航天大學計算機科學與技術學院;
【基金】:Supported by National Natural Science Foundation of China(61422204,61473149) Jiangsu Natural Science Foundation for Young Scholar(BK2013-0034) Foundation of Graduate Innovation Center in NUAA(KFJJ20151605)
【分類號】:TP311.13
[Abstract]:Most of the existing spectral clustering methods use the traditional distance measure to calculate the similarity between samples, which only considers the similarity between pairwise samples and neglects the neighboring information, and does not take into account the global distribution structure of the data. Therefore, a new spectral clustering method combining Euclidean distance and Kendall Tau distance is proposed. By combining the direct distance between two samples and the adjacent information around it, the method makes full use of the advantages of different similarity measures to capture the structural information between data from different angles. A more comprehensive reflection of the underlying structure of the data information. By comparing with the experimental results of traditional clustering algorithm on UCI standard data set, it is verified that the proposed method can significantly improve the clustering effect.
【作者單位】: 南京航空航天大學計算機科學與技術學院;
【基金】:Supported by National Natural Science Foundation of China(61422204,61473149) Jiangsu Natural Science Foundation for Young Scholar(BK2013-0034) Foundation of Graduate Innovation Center in NUAA(KFJJ20151605)
【分類號】:TP311.13
【相似文獻】
相關期刊論文 前2條
1 王佃來;劉文萍;黃心淵;;基于Sen+Mann-Kendall的北京植被變化趨勢分析[J];計算機工程與應用;2013年05期
2 ;[J];;年期
相關會議論文 前4條
1 盧學桂;;季節(jié)性Kendall檢驗在潮州供水樞紐水庫水質趨勢分析中的應用[A];中國水利學會2013學術年會論文集——S1水資源與水生態(tài)[C];2013年
2 周治年;彭長華;;Pearson、Spearman與Kendall's Tau相關分析的Excel實現(xiàn)[A];中華醫(yī)學會第八次全國檢驗醫(yī)學學術會議暨中華醫(yī)學會檢驗分會成立30周年慶典大會資料匯編[C];2009年
3 李旺君;呂昌河;;陜北1981-2011年區(qū)域干燥度的變化[A];自然地理學與生態(tài)安全學術論文摘要集[C];2012年
4 吳昊e,
本文編號:2375291
本文鏈接:http://sikaile.net/kejilunwen/ruanjiangongchenglunwen/2375291.html
最近更新
教材專著