基于加權內(nèi)容-結構網(wǎng)絡和隨機游走的社團劃分算法
發(fā)布時間:2018-07-17 05:43
【摘要】:針對傳統(tǒng)模塊優(yōu)化社團劃分算法僅能利用網(wǎng)絡的結構信息,而無法利用同樣豐富的內(nèi)容信息,導致劃分精度較低的問題,提出一種結合內(nèi)容屬性并通過給連邊加權來全面優(yōu)化網(wǎng)絡拓撲結構的社團劃分算法CCSRW(Classification with Content-Structure and Random Walk).設計利用隨機游走理論計算結構節(jié)點與內(nèi)容節(jié)點間的相似性關系矩陣,并將結構節(jié)點映射到內(nèi)容屬性空間上,最終把社團劃分問題轉化為多維無監(jiān)督聚類問題.通過在真實數(shù)據(jù)集上進行的全面實驗分析,展示了相比于傳統(tǒng)社團劃分算法,本文的算法能更準確的描述網(wǎng)絡結構,顯著提高劃分性能,并有效解決小社團不敏感問題,更適用于大規(guī)模復雜信息網(wǎng)絡的社團劃分.
[Abstract]:In view of the traditional modular optimization community partition algorithm can only use the network structure information, but can not use the same rich content information, resulting in low division accuracy problem. A community partition algorithm CCSRW (Classification with Content-Structure and Random Walk) is proposed, which combines the content attributes and weights the connected edges to optimize the topology of the network. The similarity matrix between structure node and content node is calculated by random walk theory, and the structure node is mapped to content attribute space. Finally, the problem of community division is transformed into multi-dimensional unsupervised clustering problem. Through a comprehensive experimental analysis on real data sets, it is shown that compared with the traditional community partition algorithm, the proposed algorithm can describe the network structure more accurately, significantly improve the partition performance, and effectively solve the problem of small community insensitivity. More suitable for large-scale complex information network community division.
【作者單位】: 電子科技大學計算機科學與工程學院;大眾點評網(wǎng);電子科技大學信息與軟件工程學院;
【基金】:國家科技支撐計劃(No.2013BAH33F02) 國家自然科學基金(No.61300192) 中央高;究蒲袠I(yè)務費電子科技大學項目(No.ZYGX2014J052) 2015年省科技廳支持計劃(No.2015GZ0102) 四川省自貢市公安局-基于智能視頻分析的交通流量監(jiān)控與事故預測系統(tǒng)的研究與實現(xiàn) 四川省公安廳科研項目(No.2015SCYYCX06) 成都市科學技術局軟科學研究項目(No.2015-RK00-00247-ZF)
【分類號】:TP393.02
本文編號:2129265
[Abstract]:In view of the traditional modular optimization community partition algorithm can only use the network structure information, but can not use the same rich content information, resulting in low division accuracy problem. A community partition algorithm CCSRW (Classification with Content-Structure and Random Walk) is proposed, which combines the content attributes and weights the connected edges to optimize the topology of the network. The similarity matrix between structure node and content node is calculated by random walk theory, and the structure node is mapped to content attribute space. Finally, the problem of community division is transformed into multi-dimensional unsupervised clustering problem. Through a comprehensive experimental analysis on real data sets, it is shown that compared with the traditional community partition algorithm, the proposed algorithm can describe the network structure more accurately, significantly improve the partition performance, and effectively solve the problem of small community insensitivity. More suitable for large-scale complex information network community division.
【作者單位】: 電子科技大學計算機科學與工程學院;大眾點評網(wǎng);電子科技大學信息與軟件工程學院;
【基金】:國家科技支撐計劃(No.2013BAH33F02) 國家自然科學基金(No.61300192) 中央高;究蒲袠I(yè)務費電子科技大學項目(No.ZYGX2014J052) 2015年省科技廳支持計劃(No.2015GZ0102) 四川省自貢市公安局-基于智能視頻分析的交通流量監(jiān)控與事故預測系統(tǒng)的研究與實現(xiàn) 四川省公安廳科研項目(No.2015SCYYCX06) 成都市科學技術局軟科學研究項目(No.2015-RK00-00247-ZF)
【分類號】:TP393.02
【相似文獻】
相關期刊論文 前4條
1 劉陽;季新生;劉彩霞;;網(wǎng)絡社區(qū)發(fā)現(xiàn)優(yōu)化:基于隨機游走的邊權預處理方法[J];電子與信息學報;2013年10期
2 劉輝;;基于隨機游走算法的社交網(wǎng)絡構建[J];湖北成人教育學院學報;2014年01期
3 雷鈺麗;李陽;王崇駿;劉紅星;謝俊元;;基于權重的馬爾可夫隨機游走相似度度量的實體識別方法[J];河北師范大學學報(自然科學版);2010年01期
4 杜興盛;;基于隨機游走的網(wǎng)絡故障節(jié)點定位算法仿真[J];計算機仿真;2013年06期
相關會議論文 前1條
1 燕飛;張銘;譚裕韋;唐建;鄧志鴻;;綜合社會行動者興趣和網(wǎng)絡拓撲的社區(qū)發(fā)現(xiàn)方法[A];NDBC2010第27屆中國數(shù)據(jù)庫學術會議論文集(B輯)[C];2010年
相關碩士學位論文 前1條
1 戴小玲;基于隨機游走的Web服務推薦算法研究[D];湖南科技大學;2016年
,本文編號:2129265
本文鏈接:http://sikaile.net/guanlilunwen/ydhl/2129265.html
最近更新
教材專著