【摘要】：隨著計(jì)算機(jī)技術(shù)與互聯(lián)網(wǎng)技術(shù)的迅猛發(fā)展,各類在線社會(huì)網(wǎng)絡(luò)不斷涌現(xiàn)。在線社會(huì)網(wǎng)絡(luò)具有參與自由、使用方便、信息傳播速度快、互動(dòng)性強(qiáng)等特點(diǎn),吸引了大量用戶的關(guān)注與參與,并已成為人們?nèi)粘Ｉ钆c工作中最為重要的信息交流平臺(tái)之一�；谵D(zhuǎn)發(fā)、評(píng)論等交互行為,用戶之間可以建立起虛擬的社交關(guān)系,在一定程度上可以被認(rèn)為是真實(shí)社會(huì)關(guān)系在網(wǎng)絡(luò)世界的延伸。研究在線社會(huì)網(wǎng)絡(luò)中存在的規(guī)律與現(xiàn)象對(duì)解決實(shí)際社會(huì)中相關(guān)問(wèn)題具有重要的指導(dǎo)意義。用戶作為在線社會(huì)網(wǎng)絡(luò)的核心,其特征與在線社會(huì)網(wǎng)絡(luò)拓?fù)浣Y(jié)構(gòu)的建立、用戶類型劃分、信息傳播規(guī)律等研究?jī)?nèi)容密切相關(guān)。如何有效的利用在線社會(huì)網(wǎng)絡(luò)用戶特征進(jìn)行分析研究對(duì)諸如輿情監(jiān)控、精準(zhǔn)化營(yíng)銷等多個(gè)應(yīng)用領(lǐng)域具有重大的應(yīng)用價(jià)值。本文以大規(guī)模真實(shí)在線社會(huì)網(wǎng)絡(luò)數(shù)據(jù)為對(duì)象,通過(guò)研究與分析網(wǎng)絡(luò)用戶的各類特征對(duì)在線社會(huì)網(wǎng)絡(luò)領(lǐng)域中幾類重要研究問(wèn)題展開了細(xì)致的工作。具體包括用戶影響力評(píng)估、用戶類型劃分以及基于在線社會(huì)網(wǎng)絡(luò)的人群健康研究。其中,影響力評(píng)估以及用戶類型劃分一直以來(lái)都是研究者關(guān)注的重點(diǎn),研究成果具有使輿情監(jiān)控更加高效,以及提高商業(yè)營(yíng)銷的精準(zhǔn)性等諸多價(jià)值�；谠诰�社會(huì)網(wǎng)絡(luò)的人群健康研究是近年來(lái)新興的熱點(diǎn)研究問(wèn)題,網(wǎng)絡(luò)平臺(tái)從數(shù)據(jù)到方法上可以給相關(guān)研究提供一條嶄新的問(wèn)題解決途徑,對(duì)傳統(tǒng)方法能夠起到有益的補(bǔ)充或者替代。本文主要研究?jī)?nèi)容與研究成果如下:第一,在分析用戶連接拓?fù)浣Y(jié)構(gòu)、用戶行為以及用戶信息等特征的基礎(chǔ)上,本文提出了一個(gè)基于用戶多類特征的影響力評(píng)估方法。該方法針對(duì)如何合理使用用戶屬性度量用戶影響力的問(wèn)題,首先通過(guò)貝葉斯網(wǎng)絡(luò)綜合分析各類用戶特征對(duì)用戶影響力的影響,隨后借鑒PageRank算法思想衡量鄰接節(jié)點(diǎn)的連接關(guān)系會(huì)對(duì)用戶影響力產(chǎn)生作用,從用戶自身屬性與鄰接屬性兩個(gè)方面對(duì)用戶影響力進(jìn)行評(píng)估。該方法能夠避免現(xiàn)有方法中單一指標(biāo)評(píng)估對(duì)用戶影響力反映不夠全面以及采用各屬性加權(quán)方法中屬性或行為量綱不一致所造成的物理含義不清楚等問(wèn)題。最后,基于新浪微博數(shù)據(jù)進(jìn)行了相關(guān)實(shí)驗(yàn),結(jié)果驗(yàn)證了本文方法的有效性。第二,在分析在線社會(huì)網(wǎng)絡(luò)用戶轉(zhuǎn)發(fā)鏈的基礎(chǔ)上,本文提出了一個(gè)網(wǎng)絡(luò)用戶區(qū)域交互模型。該模型以用戶行為特征為基礎(chǔ),描述了用戶與其他不同鄰接距離用戶之間的交互行為,能夠真實(shí)體現(xiàn)在線社會(huì)網(wǎng)絡(luò)用戶之間的交互模式。隨后,基于區(qū)域交互模型研究了在線社會(huì)網(wǎng)絡(luò)用戶類型劃分方法,從用戶行為與用戶影響力范疇等角度更為真實(shí)的體現(xiàn)出用戶所屬類型及其在網(wǎng)絡(luò)中所處的地位。劃分方法通過(guò)區(qū)域交互模型計(jì)算用戶對(duì)鄰接節(jié)點(diǎn)以及非鄰接節(jié)點(diǎn)的顯性、隱性影響力,根據(jù)不同類型網(wǎng)絡(luò)用戶交互模式以及兩類影響力的分布模式將網(wǎng)絡(luò)用戶劃分為重要用戶、普通用戶以及異常用戶。最后,實(shí)驗(yàn)結(jié)果表明本文提出的模型能夠針對(duì)不同類型在線社會(huì)網(wǎng)絡(luò)用戶進(jìn)行有效的識(shí)別,并且與現(xiàn)有方法相比能夠更為有效的解決相關(guān)問(wèn)題。第三,本文基于在線社會(huì)網(wǎng)絡(luò)用戶信息特征對(duì)現(xiàn)實(shí)社會(huì)中的人群健康狀況展開了探索研究,在測(cè)量分析的基礎(chǔ)上借鑒傳染病模型思想提出了一個(gè)人群肥胖預(yù)測(cè)方法。在線社會(huì)網(wǎng)絡(luò)用戶信息中包含了大量與用戶生活習(xí)慣、興趣愛好以及用戶情感、身體狀況的內(nèi)容。通過(guò)在線社會(huì)網(wǎng)絡(luò)到真實(shí)社會(huì)的映射,現(xiàn)實(shí)中人群健康狀況可以由這些用戶發(fā)布的信息內(nèi)容體現(xiàn)。在線社會(huì)網(wǎng)絡(luò)海量用戶信息數(shù)據(jù)及其開放性給人群健康相關(guān)研究提供了一個(gè)堅(jiān)實(shí)的數(shù)據(jù)平臺(tái)以及解決問(wèn)題的新途徑,與基于傳統(tǒng)醫(yī)療機(jī)構(gòu)數(shù)據(jù)研究方法相比具有更高的效率。本文以現(xiàn)有研究為基礎(chǔ)提出了幾類與肥胖相關(guān)的特征,并從在線社會(huì)網(wǎng)絡(luò)用戶中提取出相關(guān)信息,然后對(duì)每一個(gè)特征與區(qū)域人群肥胖比例的相關(guān)性進(jìn)行了分析。隨后利用與肥胖相關(guān)的特征作為肥胖人群變化的系數(shù),對(duì)不同地區(qū)人群肥胖狀況的發(fā)展趨勢(shì)進(jìn)行了預(yù)測(cè)。實(shí)驗(yàn)結(jié)果顯示,文本篩選得到的幾類特征與人群肥胖有著較為密切的關(guān)系,并且基于這些特征的預(yù)測(cè)方法也具有較為理想的有效性。第四,本文在已有研究基礎(chǔ)上開發(fā)了一個(gè)基于用戶特征的在線社會(huì)網(wǎng)絡(luò)用戶分析系統(tǒng)。該系統(tǒng)能夠有效的識(shí)別微博網(wǎng)絡(luò)中的重要用戶、異常用戶等,并且可以對(duì)不同地區(qū)用戶的肥胖健康狀況進(jìn)行分析。系統(tǒng)功能與在線社會(huì)網(wǎng)絡(luò)研究領(lǐng)域中諸如影響力評(píng)估、用戶類型劃分等研究問(wèn)題密切相關(guān),具有一定的實(shí)用價(jià)值。上述研究?jī)?nèi)容與成果體現(xiàn)出了在線社會(huì)網(wǎng)絡(luò)用戶特征的重要性,同時(shí)也體現(xiàn)出了本文的研究?jī)r(jià)值。
[Abstract]:With the rapid development of computer technology and Internet technology, various kinds of online social networks are emerging. Online social networks have the characteristics of free participation, convenient use, fast information dissemination and strong interaction, which attract a large number of users'attention and participation, and have become the most important level of information exchange in people's daily life and work. Based on the interactive behaviors of forwarding and commenting, users can establish virtual social relations, which can be regarded as the extension of real social relations in the network world to a certain extent. For the core of online social network, its characteristics are closely related to the establishment of online social network topology, the classification of users, the law of information dissemination and other research contents. In this paper, we focus on large-scale real-time online social network data, and do some detailed work on several important research issues in the field of online social network by studying and analyzing the characteristics of network users. The impact assessment and user type classification have always been the focus of researchers'attention. The research results have many values, such as making public opinion monitoring more efficient and improving the accuracy of commercial marketing. The main contents and achievements of this paper are as follows: Firstly, based on the analysis of the topological structure of user connections, user behavior and user information, a user multi-class feature is proposed in this paper. In order to solve the problem of how to use user attributes reasonably to measure user influence, this method firstly analyzes the influence of various user characteristics on user influence through Bayesian network, and then uses PageRank algorithm to measure the effect of the connection relationship between adjacent nodes on user influence. This method can avoid the problems that the single index evaluation does not reflect the user's influence comprehensively in the existing methods and the physical meanings caused by the inconsistency of attributes or behavior dimensions in each attribute weighting method are not clear. Finally, based on the Sina Weibo number Secondly, based on the analysis of the online social network user forwarding chain, this paper proposes a network user area interaction model, which describes the interaction between users and other users with different adjacent distances on the basis of user behavior characteristics. Subsequently, based on the regional interaction model, the user type partitioning method of online social network is studied. From the perspective of user behavior and user influence category, the user type and its status in the network are more truly reflected. The explicit and implicit influence of users on adjacent nodes and non-adjacent nodes are calculated. According to different types of network user interaction patterns and two types of influence distribution patterns, network users are divided into important users, ordinary users and abnormal users. Finally, the experimental results show that the proposed model can be used for different types of online communities. It can identify network users effectively and solve related problems more effectively than existing methods. Thirdly, based on the characteristics of online social network users'information, this paper explores and studies the health status of people in real society, and puts forward a population fertilizer by referring to the idea of infectious disease model on the basis of measurement and analysis. Fat prediction method. User information on online social networks contains a lot of information about users'habits, interests, emotions and physical conditions. By mapping online social networks to the real world, people's health status can be reflected in the information released by these users. Information data and its openness provide a solid data platform and a new way to solve the problem for population health-related research. It is more efficient than traditional medical institution data research methods. Based on the existing research, this paper proposes several obesity-related characteristics and extracts them from online social network users. The correlation between each feature and the obesity ratio of the population in different regions was analyzed. Then, the trend of obesity in different regions was predicted by using the obesity-related characteristics as the coefficients of obesity change. Fourthly, an online social network user analysis system based on user characteristics is developed on the basis of existing research. The system can effectively identify important users, abnormal users and so on in the microblog network. The system function is closely related to the research issues in the field of online social network research, such as impact assessment, user type classification and so on, and has certain practical value. The research value of this article is given.
【學(xué)位授予單位】：西安理工大學(xué)
【學(xué)位級(jí)別】：博士
【學(xué)位授予年份】：2016
【分類號(hào)】：TP393.09

【參考文獻(xiàn)】

相關(guān)期刊論文 前1條

1 毛佳昕;劉奕群;張敏;馬少平;;基于用戶行為的微博用戶社會(huì)影響力分析[J];計(jì)算機(jī)學(xué)報(bào);2014年04期

，

本文編號(hào)：2181744