本文選題：復(fù)雜網(wǎng)絡(luò) + 鏈路預(yù)測�。� 參考：《西安電子科技大學(xué)》2015年碩士論文

【摘要】：現(xiàn)實生活中以及科研工作中運用到的各個單位以及他們之間的關(guān)系可以抽象化成一個網(wǎng)絡(luò),由于網(wǎng)絡(luò)信息的復(fù)雜性,將這種網(wǎng)絡(luò)稱之為復(fù)雜網(wǎng)絡(luò)。復(fù)雜網(wǎng)絡(luò)就是復(fù)雜系統(tǒng)的結(jié)構(gòu),其中包括結(jié)構(gòu)復(fù)雜性:就是網(wǎng)絡(luò)系統(tǒng)具有豐富的結(jié)構(gòu)他包括社區(qū),基序,集聚性,生成規(guī)律性等。網(wǎng)絡(luò)的結(jié)構(gòu)可能會隨著時間而變化的;節(jié)點復(fù)雜性,它包括復(fù)雜網(wǎng)絡(luò)之間相互影響的復(fù)雜性以及網(wǎng)絡(luò)分層結(jié)構(gòu)的復(fù)雜性;網(wǎng)絡(luò)進化,表現(xiàn)在節(jié)點或鏈接的產(chǎn)生與消失,這也表明了網(wǎng)絡(luò)結(jié)構(gòu)的時變性;連接多樣性,他包括連接權(quán)重的多樣以及方向的多樣性;動力學(xué)復(fù)雜性以及多重復(fù)雜性融合等等。以上的種種特征表明,廣義網(wǎng)絡(luò)的復(fù)雜性可從多方面去討論研究。復(fù)雜網(wǎng)絡(luò)根據(jù)節(jié)點分布社區(qū)集聚特性,可分為單分網(wǎng)絡(luò)和二分網(wǎng)絡(luò)。復(fù)雜網(wǎng)絡(luò)中所有節(jié)點之間都存在連接關(guān)系或是存在潛在的連接關(guān)系的網(wǎng)絡(luò)稱之為單分網(wǎng)絡(luò);然而二分網(wǎng)絡(luò)是將所有的節(jié)點劃分為兩個集合,兩個集合內(nèi)部之間沒有連接關(guān)系,集合之間存在連接關(guān)系或是存在可能的連接關(guān)系。網(wǎng)絡(luò)的鏈路預(yù)測是指通過已有的節(jié)點連接關(guān)系去預(yù)測不存在連接關(guān)系的節(jié)點存在連接關(guān)系的可能性。這種預(yù)測既包含了對本身不存在且以后也不會存在鏈接的預(yù)測,同時也包含了對未來可能存在鏈接的預(yù)測。本文所做工作如下:首先了解了復(fù)雜網(wǎng)絡(luò)以及網(wǎng)絡(luò)鏈路預(yù)測的相關(guān)知識,通過生物種群網(wǎng)絡(luò)之間的互惠和捕殺行為中找到二分網(wǎng)絡(luò)在復(fù)雜網(wǎng)絡(luò)中的具體實現(xiàn),同時還發(fā)現(xiàn)在實際生活中存在著很多二分網(wǎng)絡(luò)跡象。通過對二分網(wǎng)絡(luò)特性的了解,找到關(guān)于二分網(wǎng)絡(luò)特有的鏈路預(yù)測方法,不僅僅局限于現(xiàn)有的一般性的鏈路預(yù)測方法,這種鏈路預(yù)測方法就是基于局部路徑的思想而得到的算法。首先觀察到二分網(wǎng)絡(luò)路徑長度只存在奇數(shù)路徑,因此從指數(shù)度量函數(shù)聯(lián)想到刪除偶數(shù)路徑之后就可得到奇數(shù)路徑,而這個奇數(shù)路徑從數(shù)學(xué)的角度上來看,其公式就是三角函數(shù)中的雙曲正弦函數(shù);還包括馮諾依曼指標(biāo),也是同樣進行奇數(shù)部分的保留來進行二分網(wǎng)絡(luò)的鏈路預(yù)測。通過對二分網(wǎng)絡(luò)的了解與預(yù)測,在進行路徑矩陣分析發(fā)現(xiàn),在路徑矩陣中存在著重復(fù)路徑的問題,且路徑長度越長其重復(fù)的個數(shù)越多,造成不必要的資源浪費,且在一定程度上影響著網(wǎng)絡(luò)真實路徑信息的觀察和了解,去重復(fù)路徑問題就成為本文現(xiàn)階段討論的主要問題。通過對路徑矩陣生成的形式觀察,找出重復(fù)路徑產(chǎn)生的原因,以及去重復(fù)路徑的方法。在發(fā)現(xiàn)去除重復(fù)路徑之后的預(yù)測結(jié)果能夠良好的得到預(yù)想的實驗結(jié)果。從二分網(wǎng)絡(luò)中聯(lián)想到在一般網(wǎng)絡(luò)中是否實際也同樣存在著重復(fù)路徑,答案是肯定的。但是由于網(wǎng)絡(luò)本身的性質(zhì)因此它并不區(qū)分奇數(shù)路徑和偶數(shù)路徑。采用和二分網(wǎng)絡(luò)同樣的思路進行重復(fù)路徑的去除,再對其進行實驗分析。
[Abstract]:In the real life and the scientific research work each unit and the relations between them can be abstracted into a network, because of the complexity of network information, this network is called complex network. Complex network is the structure of complex system, which includes the complexity of structure: the network system has rich structure including community, motif, agglomeration, generating regularity and so on. The structure of the network may vary over time; the complexity of nodes, which includes the complexity of interactions between complex networks and the complexity of the hierarchical structure of networks; and the evolution of networks, as shown by the generation and disappearance of nodes or links, It also shows that the network structure is time-varying; the connection diversity, which includes the diversity of connection weights and directions, the dynamic complexity and the fusion of multiple complexity, etc. The above characteristics show that the complexity of generalized networks can be discussed from many aspects. The complex network can be divided into a single network and a binary network according to the characteristics of node distributed community agglomeration. A network in which all nodes in a complex network have a connection or a potential connection is called a single network; however, a binary network divides all nodes into two sets, and there is no connection between the two sets. There is a connection relationship or a possible join relationship between sets. The link prediction of the network refers to the possibility of predicting the existence of the connection relationship between the nodes that do not exist by the existing node connection relationship. This kind of prediction includes not only the prediction that the link itself does not exist and will not exist in the future, but also the prediction of the possible existence of the link in the future. The work of this paper is as follows: firstly, we understand the related knowledge of complex network and network link prediction, and find out the concrete realization of bipartite network in complex network through the reciprocity and killing behavior between biological population networks. At the same time also found in real life there are a lot of dichotomous network signs. Through the understanding of the characteristics of binary networks, we find out that the specific link prediction methods of binary networks are not limited to the existing general link prediction methods. This link prediction method is based on the idea of local path. First, it is observed that there are only odd paths in the path length of binary network, so the odd path can be obtained by associating exponential metric function with deleting even path, and this odd path is mathematically considered. The formula is the hyperbolic sinusoidal function in trigonometric function and the von Neumann index which is also reserved for odd parts to predict the link of bipartite network. Through the understanding and prediction of the binary network, the path matrix analysis shows that there is the problem of repeated paths in the path matrix, and the longer the path length, the more the number of repeat, resulting in unnecessary waste of resources. To some extent, it affects the observation and understanding of the real path information of the network. By observing the form of path matrix generation, we find out the cause of duplicate path and the method of de-repeating path. The predicted results can get the expected experimental results well after the repetitive paths are found out. The answer is yes. However, because of the nature of the network itself, it does not distinguish odd path from even path. The same idea as the binary network is used to remove the repetitive path, and then the experimental analysis is carried out.
【學(xué)位授予單位】：西安電子科技大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2015
【分類號】：O157.5

【參考文獻】

相關(guān)期刊論文 前3條

1 汪嶺;傅秀芬;王曉牡;;基于大數(shù)據(jù)集的混合動態(tài)協(xié)同過濾算法研究[J];廣東工業(yè)大學(xué)學(xué)報;2014年03期

2 鄧智龍;淦文燕;;復(fù)雜網(wǎng)絡(luò)中的社團結(jié)構(gòu)發(fā)現(xiàn)方法[J];計算機科學(xué);2012年S1期

3 呂琳媛;;復(fù)雜網(wǎng)絡(luò)鏈路預(yù)測[J];電子科技大學(xué)學(xué)報;2010年05期

，

本文編號：2022450