本文選題：銀行網(wǎng)絡(luò) + 復(fù)雜網(wǎng)絡(luò)��； 參考：《西南財(cái)經(jīng)大學(xué)》2012年碩士論文

【摘要】：當(dāng)今世界,隨著金融創(chuàng)新不斷加快,金融體系也變得日益復(fù)雜,各種金融產(chǎn)品、金融工具與全球金融市場(chǎng)中的各種金融機(jī)構(gòu),甚至金融機(jī)構(gòu)以外的其它機(jī)構(gòu)都錯(cuò)綜復(fù)雜的聯(lián)系在一起。這意味著現(xiàn)代金融體系具有其組成部分之間相互高度鏈接的特征,這種鏈接可以用復(fù)雜網(wǎng)絡(luò)的形式形象而直觀地表現(xiàn)出來(lái),這就是所謂的金融網(wǎng)絡(luò)。 2008年的金融危機(jī)使很多國(guó)家的經(jīng)濟(jì)遭受重創(chuàng),金融危機(jī)的爆發(fā),引發(fā)人們對(duì)整個(gè)金融體系安全的思考。現(xiàn)有的研究表明,僅憑微觀層面管理的努力(微觀審慎管理)很難實(shí)現(xiàn)整個(gè)金融體系的穩(wěn)定。于是人們開始探索新的宏觀調(diào)控策略和系統(tǒng)性風(fēng)險(xiǎn)管理辦法。金融危機(jī)之后,國(guó)際金融體系改革的核心內(nèi)容之一就是大力提倡對(duì)金融體系實(shí)施宏觀審慎監(jiān)管。從這個(gè)角度來(lái)看,應(yīng)該把經(jīng)濟(jì)活動(dòng)、金融市場(chǎng)以及金融機(jī)構(gòu)的行為當(dāng)作一個(gè)整體來(lái)考慮,從全局的角度去評(píng)估和防范金融體系的風(fēng)險(xiǎn)。歐洲央行出版的《金融穩(wěn)定評(píng)論》(2010)中指出：由于金融體系中各組成部分之間相互鏈接的特性,研究現(xiàn)代金融需要引入復(fù)雜網(wǎng)絡(luò)分析方法。 從小的生物神經(jīng)網(wǎng)絡(luò)到大的電信電力網(wǎng)絡(luò),再到我們熟悉的Internet,由于它們的節(jié)點(diǎn)(結(jié)點(diǎn))龐大,鏈接關(guān)系復(fù)雜,我們稱之為復(fù)雜網(wǎng)絡(luò)。在復(fù)雜網(wǎng)絡(luò)分析方法中,由于節(jié)點(diǎn)之間鏈接的原因,對(duì)某個(gè)節(jié)點(diǎn)來(lái)說(shuō),相鄰節(jié)點(diǎn)的變化對(duì)其有重要影響,甚至?xí)艿經(jīng)]有與其直接鏈接節(jié)點(diǎn)的影響。因此,我們不能單獨(dú)去分析一個(gè)或者幾個(gè)節(jié)點(diǎn)的特征和行為,應(yīng)該把有著“關(guān)系”的所有節(jié)點(diǎn)作為一個(gè)整體來(lái)考慮。 傳統(tǒng)數(shù)據(jù)挖掘的目的是尋找海量數(shù)據(jù)中隱藏的規(guī)律,而復(fù)雜網(wǎng)絡(luò)理論的發(fā)展也是希望發(fā)現(xiàn)網(wǎng)絡(luò)中節(jié)點(diǎn)之間聯(lián)系的規(guī)律,所以從根本上來(lái)說(shuō),用復(fù)雜網(wǎng)絡(luò)的理論和方法去分析現(xiàn)實(shí)中的復(fù)雜系統(tǒng),也是一種數(shù)據(jù)挖掘模式,網(wǎng)絡(luò)化數(shù)據(jù)挖掘由此產(chǎn)生。 網(wǎng)絡(luò)化數(shù)據(jù)挖掘,就是將網(wǎng)絡(luò)拓?fù)渥鳛橐环N知識(shí)表示方式,將大規(guī)模原始數(shù)據(jù)對(duì)象及其關(guān)系抽象為網(wǎng)絡(luò)拓?fù)涞男问?采用傳統(tǒng)的數(shù)據(jù)挖掘思想,結(jié)合復(fù)雜網(wǎng)絡(luò)的理論和方法對(duì)復(fù)雜網(wǎng)絡(luò)的拓?fù)鋵傩院徒Y(jié)構(gòu)特征進(jìn)行分析和挖掘,發(fā)現(xiàn)蘊(yùn)涵在其中的、反映網(wǎng)絡(luò)中聯(lián)系規(guī)律的知識(shí)和信息。 金融網(wǎng)絡(luò)是眾多復(fù)雜網(wǎng)絡(luò)中的一種,因此可以使用網(wǎng)絡(luò)化數(shù)據(jù)挖掘的方法來(lái)分析金融網(wǎng)絡(luò)。它的優(yōu)點(diǎn)就是,可以從全局的角度去分析系統(tǒng),而不是孤立的去分析某個(gè)節(jié)點(diǎn),這就有助于對(duì)金融體系更深入的認(rèn)識(shí),對(duì)于合理布局金融基礎(chǔ)設(shè)施、預(yù)測(cè)和評(píng)估金融風(fēng)險(xiǎn)、健全金融監(jiān)管體制等方面都有著重大的意義。與歐美國(guó)家相比,我們?cè)谑褂脧?fù)雜網(wǎng)絡(luò)方法對(duì)金融市場(chǎng)、銀行系統(tǒng)、支付體系等方面的研究,做的還遠(yuǎn)遠(yuǎn)不夠(歐陽(yáng)衛(wèi)民,2010)。 銀行體系作為金融系統(tǒng)的主要組成部分,它正常高效的運(yùn)行對(duì)于一個(gè)國(guó)家金融系統(tǒng)的穩(wěn)定和健康發(fā)展意義重大。國(guó)內(nèi)外已經(jīng)有很多學(xué)者從不同的角度對(duì)銀行網(wǎng)絡(luò)作了大量的研究,包括使用大額支付系統(tǒng)信息對(duì)銀行與銀行、銀行與其它非金融機(jī)構(gòu)之間的資金流動(dòng)規(guī)律的研究,希望能夠?qū)撛诘牧鲃?dòng)性危機(jī)的進(jìn)行預(yù)警和揭示系統(tǒng)性風(fēng)險(xiǎn)的傳染規(guī)律。學(xué)者們普遍認(rèn)為,使用支付系統(tǒng)中的數(shù)據(jù)構(gòu)建銀行間資金流網(wǎng)絡(luò),最及時(shí)、最直接,也最準(zhǔn)確。 本文基于網(wǎng)絡(luò)化數(shù)據(jù)挖掘技術(shù),利用大額支付系統(tǒng)中的數(shù)據(jù),構(gòu)建銀行間資金流網(wǎng)絡(luò)(包括交易業(yè)務(wù)筆數(shù)網(wǎng)絡(luò)和業(yè)務(wù)金額網(wǎng)絡(luò))的加權(quán)復(fù)雜網(wǎng)絡(luò)拓?fù)淠Ｐ?并對(duì)其進(jìn)行了研究。包括分析了網(wǎng)絡(luò)的拓?fù)湫再|(zhì)——即兩個(gè)網(wǎng)絡(luò)的邊權(quán)和節(jié)點(diǎn)強(qiáng)度統(tǒng)計(jì)特征和分布規(guī)律；在對(duì)網(wǎng)絡(luò)結(jié)構(gòu)測(cè)度的研究發(fā)現(xiàn),網(wǎng)絡(luò)結(jié)構(gòu)屬于異配模式且整個(gè)網(wǎng)絡(luò)具有較低的聚集系數(shù)；對(duì)網(wǎng)絡(luò)進(jìn)行社區(qū)發(fā)現(xiàn)結(jié)果表明,五大國(guó)有控股商業(yè)銀行有著緊密的聯(lián)系,形成一個(gè)比較穩(wěn)定的社區(qū)。 概括來(lái)講,本文主要研究了銀行間資金流網(wǎng)絡(luò)的可視化、拓?fù)湫再|(zhì)、網(wǎng)絡(luò)結(jié)構(gòu)測(cè)度和網(wǎng)絡(luò)的社區(qū)發(fā)現(xiàn)。詳細(xì)的內(nèi)容如下。 第一部分,基于復(fù)雜網(wǎng)絡(luò)理論對(duì)銀行網(wǎng)絡(luò)研究的綜述。 首先對(duì)復(fù)雜網(wǎng)絡(luò)理論進(jìn)行了評(píng)述,包括復(fù)雜網(wǎng)絡(luò)的基本概念、統(tǒng)計(jì)特征量以及在金融銀行網(wǎng)絡(luò)上應(yīng)用的研究,然后介紹了本文重點(diǎn)研究的網(wǎng)絡(luò)結(jié)構(gòu)——加權(quán)網(wǎng)絡(luò)。接下來(lái)對(duì)使用復(fù)雜網(wǎng)絡(luò)的理論研究銀行網(wǎng)絡(luò)的現(xiàn)狀進(jìn)行了評(píng)述,同時(shí)指出了目前研究的成果和不足之處。 第二部分,對(duì)網(wǎng)絡(luò)化數(shù)據(jù)挖掘理論的綜述。 主要介紹的是本文使用的技術(shù)和理論知識(shí)。使用傳統(tǒng)數(shù)據(jù)挖掘的思想,結(jié)合復(fù)雜網(wǎng)絡(luò)的理論和方法,就是網(wǎng)絡(luò)化數(shù)據(jù)挖掘技術(shù)。在本部分評(píng)述了傳統(tǒng)的數(shù)據(jù)挖掘和網(wǎng)絡(luò)化數(shù)據(jù)挖掘的理論、模式和挖掘流程。最后對(duì)兩者的理論和方法作了比較和總結(jié)。 第三部分,對(duì)本文采用數(shù)據(jù)的描述及其可視化的研究。 本文研究數(shù)據(jù)來(lái)自中國(guó)現(xiàn)代化支付系統(tǒng)的大額支付系統(tǒng),所以首先對(duì)支付體系、支付系統(tǒng)和中國(guó)現(xiàn)代化支付系統(tǒng)作了相關(guān)評(píng)述,然后對(duì)本文使用數(shù)據(jù)的來(lái)源、整體特征作了簡(jiǎn)單的描述和初步的分析,最后使用復(fù)雜網(wǎng)絡(luò)的可視化方法分別展現(xiàn)業(yè)務(wù)金額網(wǎng)絡(luò)和業(yè)務(wù)筆數(shù)網(wǎng)絡(luò)拓?fù)鋱D型。在這部分中,還對(duì)可視技術(shù)的興起、發(fā)展和可視化算法以及可視化技術(shù)在復(fù)雜網(wǎng)絡(luò)領(lǐng)域的應(yīng)用進(jìn)行了簡(jiǎn)單評(píng)述。 第四部分,研究大額支付系統(tǒng)銀行間資金流網(wǎng)絡(luò)的拓?fù)湫再|(zhì)和結(jié)構(gòu)測(cè)度。 第一階段,對(duì)業(yè)務(wù)筆數(shù)網(wǎng)絡(luò)和業(yè)務(wù)金額網(wǎng)絡(luò)的拓?fù)湫再|(zhì)進(jìn)行分析,著重研究?jī)蓚�(gè)網(wǎng)絡(luò)的靜態(tài)統(tǒng)計(jì)特征量：邊權(quán)重和節(jié)點(diǎn)強(qiáng)度。先簡(jiǎn)單描述了邊權(quán)重和強(qiáng)點(diǎn)強(qiáng)度的分布特征,然后針對(duì)兩個(gè)網(wǎng)絡(luò)估計(jì)了它們的邊權(quán)分布的冪指數(shù)和節(jié)點(diǎn)強(qiáng)度分布的冪指數(shù)。 第二階段即網(wǎng)絡(luò)結(jié)構(gòu)測(cè)度,研究了兩個(gè)方面：網(wǎng)絡(luò)的匹配模式和聚集程度。借鑒許多學(xué)者對(duì)復(fù)雜網(wǎng)絡(luò)的研究思路,使用設(shè)置閾值的方法,分析網(wǎng)絡(luò)結(jié)構(gòu)在不同的邊權(quán)閾值下匹配模式的變化情況,然后通過(guò)計(jì)算業(yè)務(wù)金額網(wǎng)絡(luò)和業(yè)務(wù)筆數(shù)網(wǎng)絡(luò)的加權(quán)聚集系數(shù)了解整個(gè)網(wǎng)絡(luò)的聚集程度,同樣使用設(shè)置閾值的方法研究加權(quán)網(wǎng)絡(luò)的聚集系數(shù)的變化情況,并對(duì)其變化的趨勢(shì)作了相應(yīng)的分析和解釋,同時(shí)也計(jì)算了無(wú)權(quán)網(wǎng)絡(luò)的聚集系數(shù)變化情況,對(duì)兩類網(wǎng)絡(luò)進(jìn)行對(duì)比分析。 第五部分,進(jìn)一步深入挖掘銀行間資金流網(wǎng)絡(luò),研究網(wǎng)絡(luò)的社區(qū)結(jié)構(gòu)。 社區(qū)發(fā)現(xiàn)是復(fù)雜網(wǎng)絡(luò)一個(gè)重要的研究方向。首先對(duì)社區(qū)發(fā)現(xiàn)作簡(jiǎn)單評(píng)述,總結(jié)社區(qū)發(fā)現(xiàn)和傳統(tǒng)聚類分析的相似和不同之處,其次針對(duì)社區(qū)形成過(guò)程和算法物理背景對(duì)社區(qū)發(fā)現(xiàn)的研究現(xiàn)狀作簡(jiǎn)單評(píng)述,然后對(duì)本文使用的兩種社區(qū)發(fā)現(xiàn)算法：譜聚類算法和可重疊層次結(jié)構(gòu)算法的原理作了詳細(xì)介紹,并使用兩種算法對(duì)資金流網(wǎng)絡(luò)進(jìn)行社區(qū)發(fā)現(xiàn),發(fā)現(xiàn)網(wǎng)絡(luò)中隱藏的社區(qū)結(jié)構(gòu),最后對(duì)劃分結(jié)果作相應(yīng)的解釋和總結(jié)。 本文的創(chuàng)新性突出在以下兩個(gè)方面。 一是把傳統(tǒng)的數(shù)據(jù)挖掘的思想和復(fù)雜網(wǎng)絡(luò)理論結(jié)合起來(lái),研究了銀行間資金流網(wǎng)絡(luò)。全文的流程是：原始數(shù)據(jù)經(jīng)過(guò)初步預(yù)處理→構(gòu)建網(wǎng)絡(luò)拓?fù)洹鷶?shù)據(jù)展現(xiàn)(可視化)→分析網(wǎng)絡(luò)特性→發(fā)現(xiàn)結(jié)構(gòu)特征→對(duì)發(fā)現(xiàn)的結(jié)構(gòu)特征再可視化描述。運(yùn)用了一套完整的網(wǎng)絡(luò)化數(shù)據(jù)挖掘流程,對(duì)復(fù)雜網(wǎng)絡(luò)的研究和分析有更全面的認(rèn)識(shí)。 二是使用社區(qū)發(fā)現(xiàn)算法對(duì)銀行間資金流網(wǎng)絡(luò)作社區(qū)劃分。國(guó)內(nèi)外大量研究表明,社區(qū)發(fā)現(xiàn)理論已經(jīng)運(yùn)用到各種復(fù)雜網(wǎng)絡(luò)之中,許多學(xué)者從不同的方法和角度研究了銀行網(wǎng)絡(luò)的結(jié)構(gòu)和特征。但就我所知,把社區(qū)發(fā)現(xiàn)的方法運(yùn)用到銀行網(wǎng)絡(luò),特別是銀行間支付業(yè)務(wù)的資金流網(wǎng)絡(luò)中的研究是非常少的。
[Abstract]:In today ' s world , as financial innovation is accelerating , the financial system becomes increasingly complex , and financial products , financial instruments and other institutions outside the global financial market are inextricably linked . This means that modern financial systems have a highly linked feature between its components , which can be visualized in the form of complex networks , which is the so - called financial network .

Since the financial crisis in 2008 , the economy of many countries has been hit hard , and the financial crisis broke out , raising people ' s thinking about the security of the whole financial system . One of the key elements of the reform of the international financial system is to advocate the macro - prudential regulation of the financial system .

From a small biological neural network to a large telecommunications power network and to the Internet we are familiar with , because of their large nodes ( nodes ) and complex link relations , we call it a complex network . In the complex network analysis method , because of the link between nodes , the changes of adjacent nodes have important influence on them , and even be affected by no direct link nodes . Therefore , we cannot analyze the characteristics and behaviors of one or several nodes separately , and all nodes with " relation " should be considered as a whole .

The purpose of the traditional data mining is to find the hidden rules in the mass data , and the development of the complex network theory is to discover the law of the connection between nodes in the network , so the complex system in reality is analyzed by the theory and method of complex networks .

The network data mining is to use the network topology as a knowledge representation mode , abstract the large - scale raw data objects and their relationships into the form of network topology , adopt the traditional data mining thought , analyze and excavate the topological properties and structural characteristics of complex networks in combination with the theory and method of complex networks , and find out the knowledge and information in which the contact laws in the network are reflected .

The financial network is one of many complex networks , so the network data mining method can be used to analyze the financial network . It has the advantages that the system can be analyzed from the global perspective , rather than the isolated analysis of a certain node , which can help to realize the deeper understanding of the financial system , and it is far from enough for the rational distribution of financial infrastructure , forecasting and evaluation of financial risks and sound financial supervision system .

As the main component of the financial system , the banking system plays a very important role in the stability and healthy development of a country ' s financial system .

In this paper , based on the network data mining technology , the weighted complex network topology model of inter - bank fund flow network ( including transaction service pen number network and service amount network ) is constructed by using the data in large amount payment system , and the topology property of the network , i.e . , the edge weight and the node strength statistical characteristic and distribution rule of the two networks are analyzed .
In the study of network structure measure , it is found that the network structure belongs to the heterogeneous mode and the whole network has lower aggregation coefficient ;
The results of community discovery on the network show that the five state - owned commercial banks have close ties to form a more stable community .

Generally speaking , this paper mainly studies the visualization , topological property , network structure measure and community discovery of inter - bank fund flow network . The detailed contents are as follows .

The first part , based on the complex network theory to the bank network research overview .

Firstly , the complex network theory is reviewed , including the basic concept of complex network , the quantity of statistics and the research applied in the network of financial bank , then introduces the network structure _ weighting network of the research in this paper . Then the present situation of the bank network is reviewed with the theory of complex network , and the achievements and shortcomings of the current research are pointed out .

The second part is a review of the theory of networked data mining .

This part reviews the theory , model and mining flow of traditional data mining and networking data mining , and compares and summarizes the theories and methods of the two .

In the third part , the description of the data used in this paper and its visualization are studied .

This paper makes a brief description of the payment system , payment system and China ' s modern payment system , and then makes a brief description and preliminary analysis of the source and the whole characteristics of the data used in this paper . Finally , the author makes a brief comment on the rise , development and visualization algorithms of the visual technology and the application of the visualization technology in the complex network .

The fourth part is to study the topological property and the structure measure of the inter - bank fund flow network of large amount payment system .

In the first stage , the topological properties of the network and the service amount network are analyzed , and the static statistical feature quantity of the two networks is emphatically studied : the edge weight and the node strength . Firstly , the distribution characteristics of the edge weight and the strong point intensity are briefly described , and then the power index and the power exponent of the strength distribution of the nodes are estimated for the two networks .

In the second phase , network structure measure , two aspects are studied : matching mode and degree of aggregation of network .

In the fifth part , the inter - bank fund flow network is further explored , and the community structure of the network is studied .

Community discovery is an important research direction of complex network . Firstly , the author makes a brief comment on community discovery , summarizes the similarities and differences between community discovery and traditional cluster analysis , and then makes a brief comment on the research status quo of community discovery and traditional cluster analysis . Secondly , two kinds of community discovery algorithms used in this paper are introduced in detail . Two kinds of algorithms are used to find out the hidden community structure in the network , and finally explain and summarize the results of the division .

The innovative features of this article are highlighted in the following two aspects .

First , combining the traditional theory of data mining with the complex network theory , the paper studies the inter - bank capital flow network . The whole process is : the raw data is pre - processed 鈫，

本文編號(hào)：1969087