基于文本分類技術(shù)的微博平臺(tái)潛在客戶挖掘
本文選題:客戶特性描述 + 社會(huì)關(guān)系; 參考:《廣東外語(yǔ)外貿(mào)大學(xué)》2013年碩士論文
【摘要】:微博(Microblog)、Facebook和YouTube等社會(huì)化媒體的快速發(fā)展已經(jīng)深刻地改變了企業(yè)與客戶、客戶與客戶之間的溝通互動(dòng)方式,在這種新興媒體上,客戶在產(chǎn)品或服務(wù)交易市場(chǎng)上發(fā)揮著空前主動(dòng)的角色。社會(huì)化媒體具有強(qiáng)大的信息傳播能力、互動(dòng)性強(qiáng)、信息分享實(shí)時(shí)等特點(diǎn),充分利用這些特點(diǎn)進(jìn)行有效的社會(huì)化媒體營(yíng)銷能幫助企業(yè)改善品牌形象,提高品牌知名度,從而擴(kuò)大其市場(chǎng)份額。微博的用戶數(shù)量龐大、信息傳播速度迅速、影響范圍廣泛,這使得微博營(yíng)銷成為企業(yè)社會(huì)化媒體營(yíng)銷中最為重要的一個(gè)環(huán)節(jié),而潛在客戶識(shí)別是開展精準(zhǔn)微博營(yíng)銷的重要基礎(chǔ)。 如何有效地表示客戶的特性是潛在客戶挖掘最重要的基礎(chǔ)問(wèn)題,它對(duì)潛在客戶挖掘效果具有決定性的作用。目前,,國(guó)內(nèi)外對(duì)微博平臺(tái)潛在客戶挖掘的研究尚少,相關(guān)的研究主要根據(jù)客戶的人口統(tǒng)計(jì)信息和微博使用行為等方面抽取特征來(lái)刻畫客戶的特性,該類型方法的操作較為復(fù)雜;同時(shí),由于對(duì)客戶特性的描述特征還不夠準(zhǔn)確等問(wèn)題導(dǎo)致其識(shí)別準(zhǔn)確率偏低(最好的準(zhǔn)確率為76%左右)。 本研究認(rèn)為客戶的社會(huì)關(guān)系網(wǎng)的興趣愛(ài)好信息對(duì)客戶特性的描述具有重要意義,旨在通過(guò)微博平臺(tái)探索客戶的社會(huì)關(guān)系特性在潛在客戶挖掘中的作用,提出融合客戶及其微博好友自定義標(biāo)簽信息,從客戶個(gè)人和社會(huì)特性兩個(gè)方面生成客戶特性描述文本,進(jìn)而提出一種基于文本分類的微博平臺(tái)潛在客戶挖掘框架。 大量的實(shí)驗(yàn)結(jié)果表明:本研究提出的客戶特性描述方法能幫助潛在客戶識(shí)別模型平均有86%左右的準(zhǔn)確率;K最近鄰(K Nearest Neighbors,KNN)分類、樸素貝葉斯(Naive Bayes,NB)分類、Rocchio分類、基于類別質(zhì)心的分類方法(Centroid-based Classification,Centroid)和支持向量機(jī)分類(Support VectorMachines, SVM)等5種文本分類算法都獲得較高準(zhǔn)確率的潛在客戶識(shí)別效果,驗(yàn)證了本研究所提出框架的有效性。在這5個(gè)分類器中,SVM取得了準(zhǔn)確率最高的潛在客戶識(shí)別性能,但其建模和決策分析較為耗時(shí),而NB是在潛在客戶識(shí)別性能和運(yùn)行時(shí)間方面權(quán)衡的最好的分類算法,其次為Rocchio和Centroid。 借助微博平臺(tái)提供的豐富社會(huì)關(guān)系信息,融合客戶的社會(huì)關(guān)系網(wǎng)的興趣愛(ài)好信息來(lái)刻畫客戶的特性不僅為潛在客戶挖掘提供一種新的視角和手段,同時(shí)也為客戶細(xì)分、客戶流失等經(jīng)典客戶關(guān)系管理問(wèn)題的研究提供很好的參考。
[Abstract]:The rapid development of social media, such as Weibo's microblog and YouTube, has profoundly changed the way businesses communicate and interact with customers, and in this emerging media, Customers in the product or service trading market plays an unprecedented active role. Social media has the characteristics of strong information dissemination ability, strong interaction, real-time information sharing, etc. Making full use of these characteristics for effective social media marketing can help enterprises to improve their brand image and brand awareness. To expand its market share. Weibo has a large number of users, rapid information dissemination and a wide range of influence, which makes Weibo marketing become the most important part of enterprise social media marketing, and potential customer identification is an important basis for accurate Weibo marketing. How to effectively express the characteristics of customers is the most important fundamental problem in mining potential customers, which plays a decisive role in mining potential customers. At present, there are few researches on the mining of potential customers of Weibo platform at home and abroad. The related studies mainly depict the characteristics of customers according to the demographic information of customers and the characteristics of Weibo's use behavior. The operation of this type of method is more complex. At the same time, due to the inaccurate description of customer characteristics, the recognition accuracy is low (the best accuracy is about 76%). This study holds that the interest and interest information of customer social network is of great significance to the description of customer characteristics, and aims to explore the role of customer social relationship characteristics in the mining of potential customers through Weibo platform. This paper proposes a framework for potential customer mining based on text categorization based on the integration of custom tag information of clients and their Weibo friends, and the generation of customer feature description text from two aspects of customer's personal and social characteristics. A large number of experimental results show that the proposed customer characteristic description method can help potential customer identification model with an average accuracy of about 86%. K nearest neighbor K Nearest neighbor KNN, naive Bayes Bayes NB) can be used to classify Rocchio classification. The classification methods based on classification centroid (Centroid-based Classification / Centroid) and support Vector machines (SVM-based) are all effective in identifying potential customers with high accuracy, which verifies the effectiveness of the proposed framework. Among the five classifiers, SVM has achieved the highest accuracy of potential customer identification performance, but its modeling and decision analysis are time-consuming. NB is the best classification algorithm to weigh the potential customer identification performance and running time, followed by Rocchio and Centroid. With the help of the rich social relationship information provided by Weibo platform and the interest and hobby information of the social network of integrating customers to depict the characteristics of customers, it not only provides a new perspective and means for potential customers to excavate, but also provides a new way for customer segmentation. Customer drain and other classic customer relationship management issues provide a good reference.
【學(xué)位授予單位】:廣東外語(yǔ)外貿(mào)大學(xué)
【學(xué)位級(jí)別】:碩士
【學(xué)位授予年份】:2013
【分類號(hào)】:F274;G206
【參考文獻(xiàn)】
相關(guān)期刊論文 前10條
1 康健,梁允榮;分類挖掘技術(shù)在金融客戶關(guān)系管理中的應(yīng)用[J];北京理工大學(xué)學(xué)報(bào);2003年02期
2 過(guò)蓓蓓;方兆本;;基于SVM的Web日志挖掘及潛在客戶發(fā)現(xiàn)[J];管理工程學(xué)報(bào);2010年01期
3 朱宏;胡樂(lè)江;;電子商務(wù)領(lǐng)域內(nèi)潛在客戶的識(shí)別方法[J];工業(yè)工程與管理;2007年01期
4 李榮陸,胡運(yùn)發(fā);基于密度的kNN文本分類器訓(xùn)練樣本裁剪方法[J];計(jì)算機(jī)研究與發(fā)展;2004年04期
5 熊忠陽(yáng);楊營(yíng)輝;張玉芳;;基于密度的kNN分類器訓(xùn)練樣本裁剪方法的改進(jìn)[J];計(jì)算機(jī)應(yīng)用;2010年03期
6 周丹晨,殷國(guó)富,龍紅能;CRM環(huán)境下面向知識(shí)發(fā)現(xiàn)的數(shù)據(jù)分類技術(shù)的應(yīng)用研究[J];計(jì)算機(jī)應(yīng)用與軟件;2004年02期
7 徐忠海,王玲;基于客戶關(guān)系生命周期的CRM理念[J];科研管理;2003年06期
8 郭躬德;黃杰;陳黎飛;;基于KNN模型的增量學(xué)習(xí)算法[J];模式識(shí)別與人工智能;2010年05期
9 何黎;何躍;霍葉青;;微博用戶特征分析和核心用戶挖掘[J];情報(bào)理論與實(shí)踐;2011年11期
10 龐觀松;蔣盛益;;文本自動(dòng)分類技術(shù)研究綜述[J];情報(bào)理論與實(shí)踐;2012年02期
本文編號(hào):1836927
本文鏈接:http://sikaile.net/guanlilunwen/kehuguanxiguanli/1836927.html