Industry Report  JANUARY  FEBRUARY  Published by the IEEE Computer Society
Industry Report  JANUARY  FEBRUARY  Published by the IEEE Computer Society

Industry Report JANUARY FEBRUARY Published by the IEEE Computer Society - PDF document

myesha-ticknor . @myesha-ticknor
119 views | Public

Industry Report JANUARY FEBRUARY Published by the IEEE Computer Society - Description

00575132003 IEEE IEEE INTERNET COMPUTING Amazoncom Recommendations ItemtoItem Collaborative Filtering ecommendation algorithms are best known for their use on ecommerce Web sites where they use input about a cus tomers interests to generate a list of ID: 31548 Download Pdf

Tags :


Please download the presentation from below link :

Download Pdf - The PPT/PDF document "Industry Report JANUARY FEBRUARY Publ..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.




Presentation on theme: "Industry Report JANUARY FEBRUARY Published by the IEEE Computer Society"‚ÄĒ Presentation transcript

IndustryReport76JANUARY • FEBRUARY 2003 Published by the IEEE Computer Society1089-7801/03/$17.00©2003 IEEEIEEEINTERNETCOMPUTING Amazon.comRecommendations Item-to-Item Collaborative Filteringknown for their use on e-commerce Webtomer. The store radically changes based on cus-software engineer and baby toys to a new mother.important measures of Web-based and emailadvertising effectiveness — vastly exceed those ofoperate in a challenging environment. For example:•A large retailer might have huge amounts of•Many applications require the results set to be•New customers typically have extremely limit-or product ratings.•Older customers can have a glut of information, Greg Linden,Brent Smith,and Jeremy York • Amazon.com best-selling items, the algorithm typically multi-who have purchased or rated the item), making lessbased on a few customers who are most similar tothe user. It can measure the similarity of two cus-purchased it.is the number of product cat-items for each customer. However, because). Scanning every customer is approxi-), because almost all cus-regardless of the size of the catalog. But there are) processing time. Thus, the final performanceof the algorithm is approximately so, for very large data sets — such as 10 million orIt is possible to partially address these scalingWe can reduce possible to reduce the number of items examinedby a large factor.Unfortunately, all these methods also reduceto the user. Second, item-space partitioningsubject area. Third, if the algorithm discards thehave purchased only those items will not get rec-Dimensionality reduction applied to the customeralso degrade recommendation quality. To find customers who are similar to the user, clus-The segments typically are created using a clus-mined segments. Using a similarity metric, a clus-each. They then repeatedly match customers to theexisting segments, usually with some provision forvery large data sets — especially those with highreduction is also necessary. computes the user’s similarity to vectors that sum-accordingly. Some algorithms classify users into similarityABABrrrr,cos, IEEEINTERNETCOMPUTINGhttp://computer.org/internet/JANUARY • FEBRUARY 200377Amazon.com Recommendations computation is run offline. However, recommen-dation quality is low.Search-Based Methodsitems, the algorithm constructs a search query tofind other popular items by the same author,artist, or director, or with similar keywords orsubjects. If a customer buys the Godfather DVDform well. For users with thousands of purchases,however, it’s impractical to base a query on all thery of the data, reducing quality. In all cases, rec-ommendation quality is relatively poor. The rec-as best-selling drama DVD titles) or too narrowcover new, relevant, and interesting items. Popu-category fail to achieve this goal.Collaborative FilteringAmazon.com uses recommendations as a targetedmarketing tool in many email campaigns and onmost of its Web sites’ pages, including the high-Clicking on the“Your Recommendations” link leads customers to anproduct line and subject area, rate the recommendedsupermarket checkout line, but our impulse itemsare targeted to each customer.algorithms to personalize its Web site to each cus-high-quality recommendations in real time. How It Works To determine the most-similar match for a giventogether. We could build a product-to-productmatrix by iterating through all item pairs and com-puting a similarity metric for each pair. However,78JANUARY • FEBRUARY 2003 http://computer.org/internet/IEEEINTERNETCOMPUTINGIndustry Report Figure 1.The “Your Recommendations”feature on the Amazon.comhomepage.Using this feature,customers can sort recommendationsand add their own product ratings. Figure 2.Amazon.com shopping cart recommendations.The recom-mendations are based on the items in the customer’s cart:Pragmatic ProgrammerPhysics for Game Developers uct and all related products:                           use the cosine measure we described earlier, in whicheach vector corresponds to an item rather than acustomer, and the vector’s worst case. In practice, however, it’s closer totitles reduces runtime even further, with littlereduction in quality. Scalability:A ComparisonAmazon.com has more than 29 million customersWhile all this data offers opportunity, it’s also afor data sets three orders of magnitude smaller.Almost all existing algorithms were evaluated over•Traditional collaborative filtering does little orcatalog items. The algorithm is impractical onwhich reduce recommendation quality.•Cluster models can perform much of the com-is relatively poor. To improve it, it’s possible tomakes the online user–segment classification•Search-based models build keyword, category,and author indexes offline, but fail to provideThe key to item-to-item collaborative filtering’sscalability and performance is that it creates thesets. Because the algorithm recommends highlyis excellent.Unlike traditional collaborative fil-tering, the algorithm also performs well with lim-ited user data, producing high-quality recommen-Recommendation algorithms provide an effectiveform of targeted marketing by creating a person-alized shopping experience for each customer. Forlarge retailers like Amazon.com, a good recom-changes in a user’s data, and makes compellingnumber of purchases and ratings. Unlike otheralgorithms, item-to-item collaborative filtering isIn the future, we expect the retail industry totargeted marketing, both online and offline. Whilepersonalization, the technology’s increased conver-approaches will also make it compelling to offline IEEEINTERNETCOMPUTINGhttp://computer.org/internet/JANUARY • FEBRUARY 200379Amazon.com Recommendations IEEEINTERNETCOMPUTING References1.J.B. Schafer, J.A. Konstan, and J. Reidl, “E-Commerce Rec-, Kluwer Academic, 2001, pp. 115-153.2.P. Resnick et al., “GroupLens: An Open Architecture forProc. ACM 1994 Conf.Computer Supported Cooperative Work, ACM Press, 1994,3.J. Breese, D. Heckerman, and C. Kadie, “Empirical Analy-4.B.M. Sarwarm et al., “Analysis of Recommendation Algo-ACM Conf. Electronic CommerceACM Press, 2000, pp.158-167.5.K. Goldberg et al., “Eigentaste: A Constant Time Collabo-rative Filtering Algorithm,” 4, no. 2, July 2001, pp. 133-151.6.P.S. Bradley, U.M. Fayyad, and C. Reina, “Scaling Clustering, Kluwer Academic, 1998, pp. 9-15.7.L. Ungar and D. Foster, “Clustering Methods for Collabo-Proc. Workshop on Recommendation Sys-, AAAI Press, 1998.8.M. Balabanovic and Y. Shoham, “Content-Based Collabora-Comm. ACM, Mar. 1997, pp. 66-72.9.G.D. Linden, J.A. Jacobi, and E.A. Benson, US Patent 6,266,649 (to Amazon.com), Patent and Trade-mark Office, Washington, D.C., 2001.10.B.M. Sarwar et al., “Item-Based Collaborative Filtering Rec-10th Int’l World Wide Web, ACM Press, 2001, pp. 285-295. Greg Lindenwas cofounder, researcher, and senior manager inthe Amazon.com Personalization Group, where he designedsonalization, data mining, and artificial intelligence.LindenWashington. Contact him at Linden_Greg@gsb.stanford.edu. Brent Smithand an MS in mathematics from the University of Washing-ton, where he did graduate work in differential geometry. Jeremy Yorkleads the Automated Content Selection and Deliv-ery team at Amazon.com.His interests include statisticaloptimal choice of Web site display components. Hereceived a PhD in statistics from the University of Wash- ADVERTISER / PRODUCT INDEXJANUARY / FEBRUARY 2003John Wiley &SonsInside Back CoverCTIAWirelessBack Cover Advertiser / Product Page Number IEEEMedia, Advertising DirectorPhone:+1 212 419 7766Fax:+1 212 419 7589Phone:+1 714 821 8380Fax:+1 714 821 4010Email: manderson@computer.orgIEEE Computer Society,Phone:+1 714 821 8380Fax:+1 714 821 4010Phone:+1 714 821 8380Fax:+1 714 821 4010Email: dsims@computer.org Phone: +1 732 772 0160Fax: +1 732 772 0161Phone: +1 708 442 5633Fax: +1 708 442 7620Phone: +1 269 381 2156Fax: +1 269 381 2556Phone: +1 440 248 2456Fax: +1 440 248 2594Phone:+1 203 938 2418Fax:+1 203 938 3211Email: greenco@optonline.netPhone: +1 415 929 7619Fax: +1 415 577 5198Email: Phone: +1 847 705 6867Fax: +1 847 705 6878Tom WilcoxenPhone: +1 847 498 4520Fax: +1 847 498 5911Email: tw.ieeemedia@ieee.orgBarbara LynchPhone: +1 401 738 6237Fax: +1 401 739 7970Mary TononPhone: +1 415 431 5333Fax: +1 415 431 5335Tim MattesonPhone: +1 310 836 4064Fax: +1 310 836 4067German TajiriPhone:+81 42 501 9551Fax:+81 42 501 9552Email: gt.ieeemedia@ieee.orgHilary TurnbullPhone:+44 131 660 6605Fax:+44 131 660 6989Email:impress@impressmedia.comGregory MaddockPhone:+1 404 256 3800Fax:+1 404 255 7942Phone: +1 713 668 1007Fax: +1 713 668 1176Phone: +1 818 888 2407Fax: +1 818 888 4907Email: mrPhone: +1 978 244 0192Fax: +1 978 244 0103