本文選題：社交網(wǎng)絡(luò) + 移動(dòng)通信網(wǎng)絡(luò)��； 參考：《中南民族大學(xué)》2015年碩士論文

【摘要】：隨著移動(dòng)通信技術(shù)的不斷發(fā)展,越來越多的人使用手機(jī)等移動(dòng)通訊工具相互聯(lián)系。由于有些人從甲地到乙地工作或上學(xué),或者為了享受移動(dòng)通信運(yùn)營(yíng)商為新用戶提供各種優(yōu)惠,他們會(huì)經(jīng)常更換手機(jī)號(hào)碼。用戶每更換一次手機(jī)號(hào)碼,運(yùn)營(yíng)商關(guān)于該用戶的各方面信息就需要重新計(jì)算和積累,不利于運(yùn)營(yíng)商分析業(yè)務(wù)對(duì)象特征,因此如果能將多個(gè)手機(jī)號(hào)碼歸屬到同一自然人,則有利于運(yùn)營(yíng)商分析用戶使用習(xí)慣,為其提供更加個(gè)性化的服務(wù)。針對(duì)移動(dòng)通信網(wǎng)絡(luò)中同一自然人的識(shí)別問題,也就是多電話號(hào)碼歸屬同一自然人的問題,本文首先使用基于社交網(wǎng)絡(luò)的拓?fù)浣Y(jié)構(gòu)相似性的方法進(jìn)行了識(shí)別。基于社交網(wǎng)絡(luò)理論,介紹了基于節(jié)點(diǎn)局域結(jié)構(gòu)相似性、基于鏈接權(quán)重的節(jié)點(diǎn)局域相似性、融合節(jié)點(diǎn)和社團(tuán)結(jié)構(gòu)相似性、以及融合鏈接權(quán)重和社團(tuán)結(jié)構(gòu)相似性四種識(shí)別算法。在以上理論分析的基礎(chǔ)上,基于Python語言以及NetworkX和Matplotlib算法庫(kù)進(jìn)行了程序編寫、數(shù)據(jù)分析和算法實(shí)現(xiàn)。最終的實(shí)證結(jié)果表明,本文提出的基于融合鏈接權(quán)重和社團(tuán)結(jié)構(gòu)相似性的識(shí)別算法取到了很好的識(shí)別效果,具有一定的創(chuàng)新性。其次,本文提出了兩種基于信息融合的移動(dòng)通信網(wǎng)絡(luò)中同一自然人的識(shí)別算法,一種是基于參數(shù)搜索的方式,另一種是采用支持向量機(jī)的機(jī)器學(xué)習(xí)框架�；趨�(shù)搜索方法信息融合的同一自然人識(shí)別,由于考慮了消失用戶和新增用戶在屬性特征、網(wǎng)絡(luò)結(jié)構(gòu)和交互行為三個(gè)方面的綜合相似性,采用了多參數(shù)空間搜索的方法,可降低單一特征方式的不全面性和某一維度數(shù)據(jù)缺失的局限性,具有較高的準(zhǔn)確率。最后,本文提出了一種基于支持向量機(jī)的機(jī)器學(xué)習(xí)框架,融合了節(jié)點(diǎn)局域結(jié)構(gòu)的相似性、共同鄰居的加權(quán)相似性、社團(tuán)結(jié)構(gòu)相似性、節(jié)點(diǎn)的屬性相似性以及節(jié)點(diǎn)之間進(jìn)行信息交流的時(shí)間域相似性,取得了最佳的識(shí)別效果。本研究提出的算法較好解決了移動(dòng)通信網(wǎng)絡(luò)中同一自然人的識(shí)別問題,具有一定的商業(yè)價(jià)值。一方面本文的技術(shù)解決方案可以使移動(dòng)網(wǎng)絡(luò)運(yùn)營(yíng)商能更好地評(píng)估每個(gè)用戶的商業(yè)價(jià)值,也可以幫助他們估計(jì)潛在用戶的數(shù)量,制定吸引新用戶的策略。另一方面,本文方法也可以幫助運(yùn)營(yíng)商利用已有的用戶大數(shù)據(jù)來建立用戶畫像,分析用戶消費(fèi)習(xí)慣、潛在價(jià)值和忠誠(chéng)度,通過各種營(yíng)銷手段來吸引或者挽留用戶,以達(dá)到減小用戶流失、擴(kuò)大用戶規(guī)模的目的。
[Abstract]:With the development of mobile communication technology, more and more people use mobile communication tools such as mobile phones to communicate with each other.Because some people work or go to school from place A to place B, or to enjoy the benefits offered by mobile operators for new users, they often change their phone numbers.Every time a user changes a mobile phone number, the operator needs to recalculate and accumulate all aspects of information about the user, which is not conducive to the operator analyzing the characteristics of the business object. Therefore, if multiple mobile phone numbers can be assigned to the same natural person,It will be helpful for operators to analyze user usage habits and provide more personalized services for them.Aiming at the problem of identifying the same natural person in the mobile communication network, that is, the multiple telephone numbers belong to the same natural person, this paper first uses the method of topology similarity based on social network to identify the same natural person.Based on the theory of social network, this paper introduces four recognition algorithms based on node local structure similarity, link weight based node local similarity, fusion node and community structure similarity, and fusion link weight and community structure similarity.Based on the above theoretical analysis, the programming, data analysis and algorithm implementation are carried out based on Python language, NetworkX and Matplotlib algorithm library.The final empirical results show that the proposed recognition algorithm based on fusion link weight and community structure similarity has good recognition effect and is innovative to some extent.Secondly, this paper proposes two recognition algorithms for the same natural person in mobile communication networks based on information fusion, one is based on parameter search, the other is a machine learning framework based on support vector machine.For the identification of the same natural person based on the information fusion of parameter search method, considering the comprehensive similarity between vanishing user and new user in attribute feature, network structure and interactive behavior, the method of multi-parameter space search is adopted.It can reduce the incomprehensiveness of a single feature and the limitation of missing data in a certain dimension, so it has a high accuracy.Finally, a machine learning framework based on support vector machine is proposed, which combines the similarity of node local structure, the weighted similarity of common neighbor, and the similarity of community structure.The attribute similarity of nodes and the time domain similarity of information exchange between nodes have obtained the best recognition effect.The algorithm proposed in this paper solves the problem of identifying the same natural person in the mobile communication network and has certain commercial value.On the one hand, the technical solution of this paper can make mobile network operators better evaluate the commercial value of each user, and also help them estimate the number of potential users and formulate strategies to attract new users.On the other hand, this method can also help operators to use existing user big data to establish user portrait, analyze user consumption habits, potential value and loyalty, through various marketing methods to attract or retain users.In order to reduce the loss of users and expand the scale of users.
【學(xué)位授予單位】：中南民族大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2015
【分類號(hào)】：TN929.5
，

本文編號(hào)：1763505