數(shù)據(jù)挖掘在電子商務(wù)[外文翻譯]

1、<p><b>　　畢業(yè)論文外文翻譯</b></p><p><b>　　譯文</b></p><p>　　標(biāo)題：數(shù)據(jù)挖掘在電子商務(wù)</p><p>　　資料來(lái)源：網(wǎng)上科技電子圖書館 作者：大衛(wèi).李</p><p>　　摘要：現(xiàn)代商業(yè)向著電子商務(wù)發(fā)展。如果轉(zhuǎn)折適當(dāng)

2、地完成，電子商務(wù)能更好的管理企業(yè)，提供更新的服務(wù)、更低的交易成本和更好的購(gòu)買關(guān)系。電子商務(wù)成功依賴于熟練的信息技術(shù)專家，統(tǒng)計(jì)人員。這篇文章著重于統(tǒng)計(jì)人員制作幫助改變商業(yè)世界的一些貢獻(xiàn)，尤其通過(guò)開采方法的數(shù)據(jù)的發(fā)展和應(yīng)用。這是一個(gè)很大市場(chǎng)，我們采訪的主題被選擇在這個(gè)特別的問(wèn)題，為避免其他文章重疊，電子商務(wù)在提出一個(gè)很寬闊統(tǒng)計(jì)系列新的調(diào)查問(wèn)題，我們嘗試強(qiáng)調(diào)那些挑戰(zhàn)。</p><p>　　（1）介紹電子商務(wù)在改變商業(yè)的

3、形式。 </p><p>　　電子商務(wù)允許更好的客戶管理，用于營(yíng)銷的新戰(zhàn)略，一個(gè)被擴(kuò)充的產(chǎn)品系列更有效操作。這個(gè)變化的一個(gè)主要可行事物是開采工具的愈益不落俗套的數(shù)據(jù)的分布廣泛的使用。</p><p>　　商業(yè)部委托 2003 年的一項(xiàng)研究經(jīng)濟(jì)數(shù)據(jù)（美國(guó)人口調(diào)查局， 2005 年）。顯示電子商務(wù)，在百分比的基礎(chǔ)上，超過(guò)所有主要經(jīng)濟(jì)（sectorsin 2002-2003）。占制造業(yè)界銷售的

4、21.2%( 八千四百三十億美元 )（如按總數(shù)銷售圓計(jì)算）。對(duì)商人批發(fā)商領(lǐng)域，總銷售的電子商務(wù)銷售占16.9%( 七千三百億美元 ) ；占零售貿(mào)易的1.7%( 五百六十億美元 ) 和占指定的服務(wù)業(yè) 1%（五百億美元）。這些趨勢(shì)自2003年以來(lái)穩(wěn)步增長(zhǎng)。在這個(gè)領(lǐng)域的研究，有所有權(quán)性質(zhì)數(shù)據(jù)統(tǒng)計(jì)和決策科學(xué)教授和大衛(wèi)教授 .銀行研究所,杜克大學(xué),達(dá)拉謨,北卡羅萊納27708,美國(guó)(電子郵件:banks@stat.duke.edu).Yasmin

5、稱是講師,部門的應(yīng)用數(shù)學(xué)和統(tǒng)計(jì),約翰·霍普金斯大學(xué),21218年,美國(guó)在馬里蘭州的巴爾的摩算法(電子郵件:ysaid99@hotmail.com)。正在在這方面花費(fèi)了很大的努力，但大部分是秘密。當(dāng)然，亞馬遜，谷歌和微軟都深深從事統(tǒng)計(jì)研究，并在更廣泛的研究界可能了解更多有關(guān)他們的研究結(jié)果，但現(xiàn)在，這個(gè)文件能真正嘗試奠定了在相關(guān)的主要策略。在這方面，我們嘗試調(diào)查數(shù)據(jù)挖掘的貢</p><p>　　這似乎放肆討

6、論，幾乎是十年的老技術(shù)變革的歷史。但是，它是有用的檢討這段歷史，強(qiáng)調(diào)的速度和幅度電子商務(wù).在此外，重要的是要指出的技術(shù)變化的不可預(yù)見的后果是如何駕駛的研究挑戰(zhàn)在今天的應(yīng)用。</p><p><b>　　(2)客戶關(guān)系管理</b></p><p>　　新老電子商務(wù)可以做的事情之一，是客戶管理。當(dāng)?shù)晷?，社區(qū)封閉，每一個(gè)東主自動(dòng)處理客戶關(guān)系。客戶/商戶的關(guān)系變得人情味。作為

7、商業(yè)帝國(guó)的形成，電子商務(wù)現(xiàn)在允許一些回報(bào)業(yè)務(wù)的個(gè)性化服務(wù)，可以恢復(fù)客戶/商戶的關(guān)系變得人情味的可能性。</p><p>　　客戶關(guān)系管理是一個(gè)量身定做的營(yíng)銷模式并超越廣告，它包括所有的客戶體驗(yàn)方面概括：接觸，計(jì)費(fèi)，保留，幫助臺(tái)，甚至假日電子賀卡。成功的使用需要對(duì)每個(gè)客戶的詳細(xì)文件，使用數(shù)據(jù)挖掘預(yù)測(cè)特定的人想要的那種關(guān)系。</p><p>　　在這個(gè)早期的努力是基于市場(chǎng)細(xì)分。企業(yè)試圖發(fā)現(xiàn)類似

8、的消費(fèi)者群，然后制定付款計(jì)劃，廣告活動(dòng)，特別折扣優(yōu)惠，并為每個(gè)群集設(shè)計(jì)的其他政策（特別是最有利可圖的）。這個(gè)使用數(shù)據(jù)挖掘工具是聚類分析，最有名的商業(yè)先驅(qū)Claritas，其中用于人口普查數(shù)據(jù)來(lái)確定，如“孩子和時(shí)間 -囊”或“金錢的速記描述符64”集群“的消費(fèi)者，和大腦“或”回國(guó)鄉(xiāng)親“集群得到定期修訂，以反映重要的變化，例如，Claritas最近增加了集群”年輕的數(shù)字文人“，以反映重要的技術(shù)愛(ài)好者段。</p><p&g

9、t;　　客戶關(guān)系管理，可以使用這種集群建立模型。降維的方法之一是建立一個(gè)單獨(dú)的模型為每個(gè)群集;，這樣，變量，是為“背上的國(guó)家鄉(xiāng)親”，而是“年輕的數(shù)字文人無(wú)關(guān)”消費(fèi)行為顯著可吝嗇使用。為了利用這種市場(chǎng)細(xì)分的信息，分析師歸咎于客戶的群集成員從現(xiàn)有的資料，或估計(jì)在每個(gè)群集成員的概率，然后應(yīng)用貝葉斯模型平均（克萊德和喬治，2004年）或一些其他樂(lè)團(tuán)的方法“（弗里德曼和波佩斯庫(kù)，2005）。視情況而定，可用的信息可能不足以使一個(gè)群集成員的堅(jiān)強(qiáng)決心

10、。</p><p>　　第一個(gè)CRM的任務(wù)是獲得一個(gè)新客戶：超過(guò)保留當(dāng)前客戶（更昂貴的五倍，根據(jù)工商管理碩士民俗學(xué)），這通常是昂貴的，企業(yè)要招聘櫻桃挑最有利可圖的投資目標(biāo)。企業(yè)解決客戶細(xì)分的統(tǒng)計(jì)模型相結(jié)合，從cookie中的個(gè)人信息，購(gòu)買的地址列表和數(shù)據(jù)倉(cāng)庫(kù)的房屋。</p><p>　　第二個(gè)CRM的任務(wù)是為了取悅客戶部門和客戶，這可能涉及損失的領(lǐng)導(dǎo)人，幫助臺(tái)人員，24小時(shí)服務(wù)，信息技術(shù)支

11、持和開發(fā)新的網(wǎng)站服務(wù)和E -個(gè)性化（如生日卡）。然而，正如西方的邪惡女巫說(shuō)，“這些事情必須做微妙的或你的傷害法術(shù)?！庇行┛蛻艨梢酝ㄟ^(guò)周到的服務(wù)和震驚的是，他們的在線音樂(lè)供應(yīng)商知道他們的生日，其他客戶迷住了。</p><p>　　第三CRM任務(wù)是留住顧客。這是一個(gè)移動(dòng)的目標(biāo);競(jìng)爭(zhēng)對(duì)手都在不斷推出新的服務(wù)和價(jià)格。數(shù)據(jù)挖掘是一個(gè)強(qiáng)大的資產(chǎn)保留問(wèn)題，遠(yuǎn)遠(yuǎn)超過(guò)一般顧客慣性優(yōu)勢(shì)。大多數(shù)企業(yè)都對(duì)他們的客戶的詳細(xì)文件，表明所有的

12、用戶帳戶信息，加上已經(jīng)獲得的任何其他數(shù)據(jù)。顯然，這些記錄確定客戶的手機(jī)合同，必須更新或信用卡即將到期，這些是那些最有可能流失（轉(zhuǎn)）一個(gè)競(jìng)爭(zhēng)對(duì)手。這些記錄也可以標(biāo)志人的消費(fèi)習(xí)慣正在發(fā)生變化（一動(dòng)，一個(gè)新的工作，為漫畫書收集的懷舊激情）。數(shù)據(jù)挖掘可以解釋這種變化在一定程度上，它允許公司請(qǐng)客戶提供新的服務(wù)，或?yàn)樗麄兡壳暗那闆r下設(shè)計(jì)的付款計(jì)劃。</p><p>　　這是很容易，可以想象，電子商務(wù)有一天會(huì)導(dǎo)致每一筆交易成倍

13、地增長(zhǎng)Priceline.com是已經(jīng)在朝著這個(gè)方向努力。在這一天結(jié)束，我們可能會(huì)看到一個(gè)商業(yè)社會(huì)中，所有的利潤(rùn)微薄，每一筆交易都有一個(gè)特定的成本。</p><p>　　這些問(wèn)題的分析是很重要的。電子商務(wù)客戶關(guān)系管理系統(tǒng)的數(shù)據(jù)挖掘?yàn)榻y(tǒng)計(jì)人員和決策者提供了一個(gè)不尋常的設(shè)置，是豐富的數(shù)??據(jù)庫(kù)，但合理的假設(shè)，建模，如獨(dú)立的觀察是必須的。我們面臨的挑戰(zhàn)是豐富的聽話簡(jiǎn)單的數(shù)據(jù)，但仍然支持業(yè)務(wù)決策的解釋說(shuō)明提取，這是電子商

14、務(wù)的擴(kuò)展。因?yàn)樵絹?lái)越多的企業(yè)加入了電子服務(wù)，更多的人在購(gòu)物時(shí)上線，，使任何分析必須跟蹤一個(gè)移動(dòng)的目標(biāo)變得更加困難。</p><p><b>　　外文文獻(xiàn)原文</b></p><p>　　Title：Data Mining in Electronic Commerce</p><p>　　Material Source: Scientific E

15、lectronic Library Online Author: David L </p><p>　　Abstract. Modern business is rushing toward e-commerce. If the transition is done properly, it enables better management, new services, lower transac-

16、tion costs and better customer relations. Success depends on skilled information technologists, among whom are statisticians. This paper focuses on some of the contributions that statisticians are making to help change t

17、he business world, especially through the development and application of data mining methods. This is a very large area, and the t</p><p>　　1. INTRODUCTION</p><p>　　Electronic commerce is changi

18、ng the face of business. It allows better customer management, newstrategies for marketing, an expanded range of products and more efficient operations. A key enabler of this change is the widespread use of increasingly

19、sophisticated data mining tools.</p><p>　　The Department of Commerce commissioned a study of 2003 economic data (U.S. Census Bureau,2005). It showed that e-commerce, on a percentage basis, outperformed all f

20、our major economic sectorsin 2002–2003. For the manufacturing sector, 21.2%($843 billion) of the activity (as measured in total sales dollars) was classified as e-commerce. For the merchant wholesalers sector, e-commerce

21、 sales were16.9% ($730 billion) of total sales; for retail trade itwas 1.7% ($56 billion) and for selected service </p><p>　　Mathematics and Statistics, Johns Hopkins University,Baltimore, Maryland 21218, US

22、A (e-mail:ysaid99@hotmail.com).the algorithms. A great deal of effort is being expended in this area, but most of it is secret. Certainly Amazon,Google and Microsoft are deeply engaged in statistical research, and in tim

23、e the broader research commu-nity may learn more about their findings, but for now,all this paper can really attempt is to lay out the main strategies in the relevant are as.In that context, we try </p><p>　

24、　2. CUSTOMER RELATIONSHIP MANAGEMENT</p><p>　　One of the new old things that e-commerce can do is customer management. When shops were small and communities were insular, each proprietor handled customer rel

25、ations automatically. As mercantile empires formed, the customer/merchant relationship became impersonal. E-commerce now allows the possibility of recovering some of the individualized service that can cement return busi

26、ness.</p><p>　　Customer relationship management is a generalization of the tailored marketing discussed in Section 2.2.It goes beyond advertising to include all aspects of the customer experience: contact, b

27、illing, retention, help desks and even holiday e-cards. Successful use requires detailed files on each customer; one uses data mining to anticipate the kind of relationship that specific people want.</p><p>

28、　　Early efforts at this were based on market segmentation. Businesses attempted to discover clusters of consumers who were similar, and then would develop payment plans, ad campaigns, special discounts and other policies

29、 designed for each cluster (especially the most profitable). The data mining tool used for this was cluster analysis, and the most famous commercial pioneer is Claritas, which used Census data to identify 64 “clusters” o

30、f consumers, with shorthand descriptors such as “kids and cul-</p><p>　　Customer relationship management can use such clusters to build models. One approach to dimension reduction is to build a separate mode

31、l for each cluster;in this way, variables that are significant for the consumer behavior of “back country folks” but irrelevant to the “young digerati” can be parsimoniously used. To use this kind of market segmentation

32、information, the analyst either has to impute the cluster membership of a customer from available information or has to estimate the probabilitie</p><p>　　The first CRM task is to acquire a new customer:this

33、 is usually more expensive than retaining a current customer (five times more expensive, according to MBA folklore), and businesses want to target their recruitment investment to cherry-pick the most profitable ones. Bus

34、inesses address this by combining statistical models of customer segments with individual information from cookies, purchased address lists and data ware houses.</p><p>　　The second CRM task is to please the

35、 customer; depending on the sector and the kind of customer, this may involve loss leaders, help-desk personnel, 24-hour service, information technology support and development for new web site services, and e-personaliz

36、ation(such as birthday cards). However, as the Wicked Witch of the West says, “These things must be done delicately or you hurt the spell.” Some customers can be put off by overattentive service and alarmed that their on

37、-line music provider kno</p><p>　　The third CRM task is to retain customers. This is a moving target; competitors are constantly offering new services and prices. Data mining is a strong asset for the retent

38、ion problem, over and above the usual advantage of customer inertia. Most businesses have detailed files on their customers that indicate all the user-account information, plus whatever other data have been acquired. Obv

39、iously, these records identify customers whose cell phone contract must be renewed or whose credit card is </p><p>　　It is easily conceivable that electronic commerce will someday lead to microhaggling on ev

40、ery transaction;Priceline.com is already moving in this direction. At the end of the day, we may see a business community in which all margins are razor thin and every transactionhas a specific cost.</p><p>

41、　　The analysis of such questions is important. Electronic commerce provides an unusual setting for statisticians and decision-makers: it is rich in data, but poorin plausible assumptions that are useful for modeling,such

42、 as independent observations. Our challenge is touse the abundance of data in place of tractable simplicity, but still extract interpretable descriptions that support business decisions. This is made more difficult by th