scispace - formally typeset
Search or ask a question

Showing papers on "Recommender system published in 2001"


Proceedings ArticleDOI
01 Apr 2001
TL;DR: This paper analyzes item-based collaborative ltering techniques and suggests that item- based algorithms provide dramatically better performance than user-based algorithms, while at the same time providing better quality than the best available userbased algorithms.
Abstract: Recommender systems apply knowledge discovery techniques to the problem of making personalized recommendations for information, products or services during a live interaction. These systems, especially the k-nearest neighbor collaborative ltering based ones, are achieving widespread success on the Web. The tremendous growth in the amount of available information and the number of visitors to Web sites in recent years poses some key challenges for recommender systems. These are: producing high quality recommendations, performing many recommendations per second for millions of users and items and achieving high coverage in the face of data sparsity. In traditional collaborative ltering systems the amount of work increases with the number of participants in the system. New recommender system technologies are needed that can quickly produce high quality recommendations, even for very large-scale problems. To address these issues we have explored item-based collaborative ltering techniques. Item-based techniques rst analyze the user-item matrix to identify relationships between di erent items, and then use these relationships to indirectly compute recommendations for users. In this paper we analyze di erent item-based recommendation generation algorithms. We look into di erent techniques for computing item-item similarities (e.g., item-item correlation vs. cosine similarities between item vectors) and di erent techniques for obtaining recommendations from them (e.g., weighted sum vs. regression model). Finally, we experimentally evaluate our results and compare them to the basic k-nearest neighbor approach. Our experiments suggest that item-based algorithms provide dramatically better performance than user-based algorithms, while at the same time providing better quality than the best available userbased algorithms.

8,634 citations


Journal ArticleDOI
TL;DR: An explanation of how recommender systems are related to some traditional database analysis techniques is presented, and a taxonomy ofRecommender systems is created, including the inputs required from the consumers, the additional knowledge required from a database, the ways the recommendations are presented to consumers,The technologies used to create the recommendations, and the level of personalization of the recommendations.
Abstract: i>Recommender systems are being used by an ever-increasing number of E-commerce sites to help consumers find products to purchase. What started as a novelty has turned into a serious business tool. Recommender systems use product knowledge—either hand-coded knowledge provided by experts or “mined” knowledge learned from the behavior of consumers—to guide consumers through the often-overwhelming task of locating products they will like. In this article we present an explanation of how recommender systems are related to some traditional database analysis techniques. We examine how recommender systems help E-commerce sites increase sales and analyze the recommender systems at six market-leading sites. Based on these examples, we create a taxonomy of recommender systems, including the inputs required from the consumers, the additional knowledge required from the database, the ways the recommendations are presented to consumers, the technologies used to create the recommendations, and the level of personalization of the recommendations. We identify five commonly used E-commerce recommender application models, describe several open research problems in the field of recommender systems, and examine privacy implications of recommender systems technology.

1,672 citations


Proceedings ArticleDOI
01 Jan 2001
TL;DR: It was found that the time spent on a pages, the amount of scrolling on a page and the combination of time and scrolling had a strong correlation with explicit interest, while individual scrolling methods and mouse-clicks were ineffective in predicting explicit interest.
Abstract: Recommender systems provide personalized suggestions about items that users will find interesting. Typically, recommender systems require a user interface that can ``intelligently'' determine the interest of a user and use this information to make suggestions. The common solution, ``explicit ratings'', where users tell the system what they think about a piece of information, is well-understood and fairly precise. However, having to stop to enter explicit ratings can alter normal patterns of browsing and reading. A more ``intelligent'' method is to useimplicit ratings, where a rating is obtained by a method other than obtaining it directly from the user. These implicit interest indicators have obvious advantages, including removing the cost of the user rating, and that every user interaction with the system can contribute to an implicit rating.Current recommender systems mostly do not use implicit ratings, nor is the ability of implicit ratings to predict actual user interest well-understood. This research studies the correlation between various implicit ratings and the explicit rating for a single Web page. A Web browser was developed to record the user's actions (implicit ratings) and the explicit rating of a page. Actions included mouse clicks, mouse movement, scrolling and elapsed time. This browser was used by over 80 people that browsed more than 2500 Web pages.Using the data collected by the browser, the individual implicit ratings and some combinations of implicit ratings were analyzed and compared with the explicit rating. We found that the time spent on a page, the amount of scrolling on a page and the combination of time and scrolling had a strong correlation with explicit interest, while individual scrolling methods and mouse-clicks were ineffective in predicting explicit interest.

768 citations


Proceedings ArticleDOI
05 Oct 2001
TL;DR: The experimental evaluation on five different datasets show that the proposed item-based algorithms are up to 28 times faster than the traditional user-neighborhood based recommender systems and provide recommendations whose quality is up to 27% better.
Abstract: The explosive growth of the world-wide-web and the emergence of e-commerce has led to the development of recommender systems---a personalized information filtering technology used to identify a set of N items that will be of interest to a certain user. User-based Collaborative filtering is the most successful technology for building recommender systems to date, and is extensively used in many commercial recommender systems. Unfortunately, the computational complexity of these methods grows linearly with the number of customers that in typical commercial applications can grow to be several millions. To address these scalability concerns item-based recommendation techniques have been developed that analyze the user-item matrix to identify relations between the different items, and use these relations to compute the list of recommendations.In this paper we present one such class of item-based recommendation algorithms that first determine the similarities between the various items and then used them to identify the set of items to be recommended. The key steps in this class of algorithms are (i) the method used to compute the similarity between the items, and (ii) the method used to combine these similarities in order to compute the similarity between a basket of items and a candidate recommender item. Our experimental evaluation on five different datasets show that the proposed item-based algorithms are up to 28 times faster than the traditional user-neighborhood based recommender systems and provide recommendations whose quality is up to 27% better.

608 citations


Book ChapterDOI
16 Sep 2001
TL;DR: An analysis of the primary design issues for group recommenders, including questions about the nature of groups, the rights of group members, social value functions for groups, and interfaces for displaying group recommendations.
Abstract: We present PolyLens, a new collaborative filtering recommender system designed to recommend items for groups of users, rather than for individuals. A group recommender is more appropriate and useful for domains in which several people participate in a single activity, as is often the case with movies and restaurants. We present an analysis of the primary design issues for group recommenders, including questions about the nature of groups, the rights of group members, social value functions for groups, and interfaces for displaying group recommendations. We then report on our PolyLens prototype and the lessons we learned from usage logs and surveys from a nine-month trial that included 819 users We found that users not only valued group recommendations, but were willing to yield some privacy to get the benefits of group recommendations Users valued an extension to the group recommender system that enabled them to invite non-members to participate, via email

580 citations


Proceedings Article
02 Aug 2001
TL;DR: It is shown that secondary content information can often be used to overcome sparsity and appropriate mixture models incorporating secondary data produce significantly better quality recommenders than k-nearest neighbors (k-NN).
Abstract: Recommender systems leverage product and community information to target products to consumers. Researchers have developed collaborative recommenders, content-based recommenders, and a few hybrid systems. We propose a unified probabilistic framework for merging collaborative and content-based recommendations. We extend Hofmarm's (1999) aspect model to incorporate three-way co-occurrence data among users, items, and item content. The relative influence of collaboration data versus content data is not imposed as an exogenous parameter, but rather emerges naturally from the given data sources. However, global probabilistic models coupled with standard EM learning algorithms tend to drastically overfit in the sparsedata situations typical of recommendation applications. We show that secondary content information can often be used to overcome sparsity. Experiments on data from the Researchlndex library of Computer Science publications show that appropriate mixture models incorporating secondary data produce significantly better quality recommenders than k-nearest neighbors (k-NN). Global probabilistic models also allow more general inferences than local methods like k-NN.

518 citations


01 Jan 2001
TL;DR: The hypothesis was that friends would make superior recommendations since they know the user well, and have intimate knowledge of his / her tastes in a number of domains, in contrast to RS, which only have limited, domain-specific knowledge about the users.
Abstract: The quality of recommendations and usability of six online recommender systems (RS) was examined. Three book RS (Amazon.com, RatingZone & Sleeper) and three movie RS (Amazon.com, MovieCritic, Reel.com) were evaluated. Quality of recommendations was explored by comparing recommendations made by RS to recommendations made by the user’s friends. Results showed that the user’s friends consistently provided better recommendations than RS. However, users did find items recommended by online RS useful: recommended items were often “new” and “unexpected”, while the items recommended by friends mostly served as reminders of previously identified interests. Usability evaluation of the RS showed that users did not mind providing more input to the system in order to get better recommendations. Also users trusted a system more if it recommended items that they had previously liked. A common way for people to decide what books to read is to ask friends and acquaintances for recommendations. The logic behind this time-tested method is that one shares tastes in books, movies, music etc., with one’s friends. As such, items that appeal to them (friends) might appeal to me. Online Recommender Systems (RS) attempt to create a technological proxy for this social filtering process. The assumption behind many RS is that a good way to personalize recommendations for a user is to identify people with similar interests and recommend items that have interested these like-minded people (Resnick & Varian (1997), Goldberg, Nichols, Oki & Terry (1992)). This premise forms the statistical basis of most collaborative filtering algorithms. Since the goal of most RS is to replace (or at least augment) what is essentially a social process, we decided to directly compare the two ways of receiving recommendations (friends & online RS). Do users like receiving recommendations from an online system? How do the recommendations provided by online systems differ from those provided by the users’ friends? Our hypothesis was that friends would make superior recommendations since they know the user well, and have intimate knowledge of his / her tastes in a number of domains. In contrast, RS only have limited, domain-specific knowledge about the users. Also, information retrieval systems do not yet match the sophistication of human judgment processes.

487 citations


Patent
Don Turnbull1, Hinrich Schuetze1
26 Jan 2001
TL;DR: In this paper, a search and recommendation system employs the preferences and profiles of individual users and groups within a community of users, as well as information derived from categorically organized content pointers, to augment electronic commerce related searches, re-rank search results, and provide recommendations for commerce related objects based on an initial subject-matter query and an interaction history of a user.
Abstract: A search and recommendation system employs the preferences and profiles of individual users and groups within a community of users, as well as information derived from categorically organized content pointers, to augment electronic commerce related searches, re-rank search results, and provide recommendations for commerce related objects based on an initial subject-matter query and an interaction history of a user. The search and recommendation system operates in the context of a content pointer manager, which stores individual users' content pointers (some of which may be published or shared for group use) on a centralized content pointer database connected to a network. The shared content pointer manager is implemented as a distributed program, portions of which operate on users' terminals and other portions of which operate on the centralized content pointer database. A user's content pointers are organized in accordance with a local topical categorical hierarchy. The hierarchical organization is used to define a relevance context within which returned objects are evaluated and ordered.

454 citations


Patent
17 Oct 2001
TL;DR: In this paper, the authors present a set of products that they predict the consumer will prefer and/or perform well for the problem or concern identified by the consumer, and the performance and preference predictions are a function of consumer problems and product responsiveness patterns.
Abstract: Systems and methods of utilizing communications networks and multivariate analysis to predict or recommend optimal products from a predefined population of commercially available products are disclosed. The recommendations are based on intelligence contained in processing elements and subjective and/or objective product information received from consumers or input to the systems as part of their initial setup. The output of the systems comprise sets of products that they predict the consumer will prefer and/or perform well for the problem or concern identified by the consumer. The performance and preference predictions are a function of consumer problems and product responsiveness patterns. Objective product information is generally obtained with diagnostic instruments. Data measured with the diagnostic instruments may be communicated to the data processing portions of the invention via the Internet. The outputs of the data processing portion of the system may be presented to consumers via the Internet as well.

437 citations


01 Jan 2001
TL;DR: This work presents a framework for understanding recommender systems and surveys a number of distinct approaches in terms of this framework, and suggests two main research challenges: helping people form communities of interest while respecting personal privacy and developing algorithms that combine multiple types of information to compute recommendations.
Abstract: The Internet and World Wide Web have brought us into a world of endless possibilities: interactive Web sites to experience, music to listen to, conversations to participate in, and every conceivable consumer item to order. But this world also is one of endless choice: how can we select from a huge universe of items of widely varying quality? Computational recommender systems have emerged to address this issue. They enable people to share their opinions and benefit from each other’s experience. We present a framework for understanding recommender systems and survey a number of distinct approaches in terms of this framework. We also suggest two main research challenges: (1) helping people form communities of interest while respecting personal privacy, and (2) developing algorithms that combine multiple types of information to compute recommendations.

308 citations


Journal ArticleDOI
TL;DR: A personalized recommender system designed to suggest new products to supermarket shoppers in a pervasive computing environment in which supermarket customers use Personal Digital Assistants to compose and transmit their orders to the store, which assembles them for subsequent pickup.
Abstract: We describe a personalized recommender system designed to suggest new products to supermarket shoppers. The recommender functions in a pervasive computing environment, namely, a remote shopping system in which supermarket customers use Personal Digital Assistants (PDAs) to compose and transmit their orders to the store, which assembles them for subsequent pickup. The recommender is meant to provide an alternative source of new ideas for customers who now visit the store less frequently. Recommendations are generated by matching products to customers based on the expected appeal of the product and the previous spending of the customer. Associations mining in the product domain is used to determine relationships among product classes for use in characterizing the appeal of individual products. Clustering in the customer domain is used to identify groups of shoppers with similar spending histories. Cluster-specific lists of popular products are then used as input to the matching process. The recommender is currently being used in a pilot program with several hundred customers. Analysis of results to date have shown a 1.8% boost in program revenue as a result of purchases made directly from the list of recommended products. A substantial fraction of the accepted recommendations are from product classes new to the customer, indicating a degree of willingness to expand beyond present purchase patterns in response to reasonable suggestions.


Proceedings ArticleDOI
22 Oct 2001
TL;DR: In this article, the authors explore the acquisition of user profiles by unobtrusive monitoring of browsing behaviour and application of supervised machine-learning techniques coupled with an ontological representation to extract user preferences.
Abstract: Tools for filtering the World Wide Web exist, but they are hampered by the difficulty of capturing user preferences in such a dynamic environment. We explore the acquisition of user profiles by unobtrusive monitoring of browsing behaviour and application of supervised machine-learning techniques coupled with an ontological representation to extract user preferences. A multi-class approach to paper classification is used, allowing the paper topic taxonomy to be utilised during profile construction. The Quickstep recommender system is presented and two empirical studies evaluate it in a real work setting, measuring the effectiveness of using a hierarchical topic ontology compared with an extendable flat list.

Journal ArticleDOI
TL;DR: Recommender system users who rate items across disjoint domains face a privacy risk analogous to the one that occurs with statistical database queries.
Abstract: Recommender system users who rate items across disjoint domains face a privacy risk analogous to the one that occurs with statistical database queries.

Patent
30 May 2001
TL;DR: In this article, a de-centralized, or distributed, monitoring system provides for data collection across a broad range of remote sources, collecting explicit data (which may be input directly by a user in the form of recommendation, comment, or vote) and/or implicit data (according to the user's browsing activity).
Abstract: A de-centralized, or distributed, monitoring system provides for data collection across a broad range of remote sources, collecting explicit data (which may be input directly by a user in the form of recommendation, comment, or vote) and/or implicit data (which may be collected by the system according to the user's browsing activity). Data may be monitored locally at the client side, and subsequently transmitted to a central database. Data may be aggregated at the server, having been collected on the client side from multiple remote sources. During the aggregation process, data collected by the distributed monitoring system are categorized and organized in a central database for convenient retrieval. Implementation of the collected data includes both transmitting explicit data on demand as well as utilizing explicit data, implicit data, or a combination of both explicit and implicit data, in an open recommendation system which facilitates customization and personalization of the information retrieval process. A user may be provided with the option of turning off, or 'deselecting,' the implicit data collection functionality of the system.

Proceedings ArticleDOI
05 Oct 2001
TL;DR: The Music Recommendation System (MRS) is designed to provide a personalized service of music recommendation, which is based on the favorite degrees of the users to the music groups.
Abstract: With the growth of the World Wide Web, a large amount of music data is available on the Internet. In addition to searching expected music objects for users, it becomes necessary to develop a recommendation service. In this paper, we design the Music Recommendation System (MRS) to provide a personalized service of music recommendation. The music objects of MIDI format are first analyzed. For each polyphonic music object, the representative track is first determined, and then six features are extracted from this track. According to the features, the music objects are properly grouped. For users, the access histories are analyzed to derive user interests. The content-based, collaborative and statistics-based recommendation methods are proposed, which are based on the favorite degrees of the users to the music groups. A series of experiments are carried out to show that our approach is feasible.

Proceedings Article
02 Aug 2001
TL;DR: This work describes two families of methods for transforming data to encode time order in ways amenable to off-the-shelf classification and density estimation tools and compares the results to collaborative filtering without ordering information.
Abstract: We treat collaborative filtering as a univariate time series problem: given a user's previous votes, predict the next vote. We describe two families of methods for transforming data to encode time order in ways amenable to off-the-shelf classification and density estimation tools. Using a decision-tree learning tool and two real-world data sets, we compare the results of these approaches to the results of collaborative filtering without ordering information. The improvements in both predictive accuracy and in recommendation quality that we realize advocate the use of predictive algorithms exploiting the temporal order of data.

01 Jan 2001
TL;DR: The author reveals that the author’s motivation for writing the book came from a combination of personal experience and a need to know what to expect when it came to dealing with the natural disasters.
Abstract: *$ % + ( , * ( *$% -# . 0/1 23 $% 4 657 8$% % 9: /1 $% $.;<2 . ' = $%> ( ?$% A@ % ' ( $B CD 8$% E FG $% % *$HFI E# . '"KJL> G23 $%>M $%> ( N> G$%> 8 K 'CH ( $ 9 8; # 8 O$%> OFP ($% Q@( % ( ( ?9 R % $% 7 R . 8$%# $% "0ST %@3 $% 9U @3 $% MFI % V23 $%>W $%> X> (2 % Y % + ( Z * *$% [ ' \ % 6$%> . ] .> .$.; 9 "HST ^$%> _@+ @3 ' C1 R@( % . $_ U 9 $M + 0 `a $% Fb C1 %cXFd ^ ,2 9W $% $Z e 23 $% N"gf!#( @ @( % 4 >X# . , U $% $.;h2+ . ' i@ % ' ( $% R$% : >+ &( *$.; 9W# . $ ( H + j$%> X@ % ,@3 % . k ' Y .# 9 9 *$% $%>( % # 9 > 23 $% 4 Bl+ 8$% % 9("mJi H@ % . $n &(@3 % $ ( % ; .# 8$% n$%>+ $1 .> 'Co> 'Ce$%> 1 @ @( % 4 >N KpGq r+sut r sPvdw_q'q x sut.y:pGq4zdz8{ | v q }*{4sd~d t_€~Pz8sut } ~dr4‚ @3 .Fd % n23 $.$% n$%> Q@ # % H $% $.;h2 . ' @( % ' $% ' m@ #( % A 23 $% 4 l+ 8$% ' 7 i ^ + 4 > 2( % ] @ ; @( % 4 >ƒ"„Ji : . … . # . ^ $%> $% X @( % 4 0$%> U@3 .FI .; ! FG #( H> 2( % = * ( *$% 0" 1. INTRODUCTION † + ( B * ( *$% _> @U 4 % (Fd % $% 6 4 % 4 2 M@( % ( ( 9H@3 % . k ' R .# 9 9 *$% m2 . ' M , H> *$% . Q F # . '‡ ? c4 ? + ( . c4 " ) A (;h _ *$% % 1@( % H % A; + 9Q . % 1 " 9("nˆ_ k N /1‰QŠMfEJ‹ Œ1 % %ˆ_ + (Š! 2 S ) ‰Q2N $% "n?> % O % M$TC1 A@ % $B @ @( % 4 > B$% A2 # 9 % Q * *$% _Ž/1 b 23 $% =57 8$% % 9Wh/?5‘! + /1 $% $.;<2 . ' Lh/?ŒH‘M % + 9 "i/?5’ * ( *$% RC1 %cj2 $% 9A# . 1FI ' (2+ c $%> HFd % “ Fƒ $% 9 1Fd H $% 1 Z 9 4 Z Z + A & @ 8$? . % 8$% H Z ( 8`a % ? 9 @( % l FH . 4 n# . % O j $% % 9^> 'C”$% = % + X $% 0"if! j$%> Z $%> ,>+ + ƒ G $% $.;h2+ . i $%> ( R@( % ; ? % $% G2 @ % 9Q % @( % . $ $%  Fa ; $% $? $ ' Z = Z 8$% •$% % @ % . $ $% E F7 $% $1$%>+ $ $% % *$% E$%> M# . '" /1 $% $.;<2 . ' $%> ( 7 ' ,# – # 8 O > $% % k E ' >R# . ' 2 # $m/?50 *$% > 7 . 1c4 M $ 9 7 4 K$%> “—8˜ ˜ ™h"m57 % *$% 8 /?56 ' A@3 .Fd % g , mCH> % ?$%> % E  $G R# >R $% $ . . $% ' jCH 8$%>i 8$% m RCH> % $%> $% $R 8š # $OFI 6 @ # $% R$% : 8 (k ZŽ› ( ' n @ , $% "Xœ + 8 i /?5L * ( *$% > M$%> 2 8$T :$% ^@ % . % + @ $% # M % A; + $% ƒ P" "R $O 6 % + 0 8$% !$%> $O % , % $ $% -$%> Q# . ' (2 #($H $H $ = $% $?Fb % “$%> Q# . '‡ E@( % ; l " Œ1 ' # . F1$%> . , % ' . G/?5L * ( *$% Q>+ ' 4 R23 :# . ' Fd 8 % 8 j .# . *Fd# j$% U2 # W % + ( O * ( *$% R X ( % ( M— ž( ƒ˜'ž ™u"Ÿ! 'C1 4 E$%> .# `a EFb % “$ i$%> X# . .;< 8$% ¥ $% 9\ $. % 8&¦ 6$T (@ 8 ‹ 4 . .@+ % . "Y?> % FI % ^$%> ^@ % 2+ 2 $< X FBl+ 9: ] . $ F # . % GCH 8$%> § ̈u© aa ̈<« ̈u¬ ­ aa®* ̄I°R . n $% 9 n # .# 8 , 'CM" ?> Q ! FI$% 6$%> O ' . RCH> : * ( *$% _>+ ' 4 4 . U> 9 > $% A;h$% ;h# . $% "i?> ,@( % 2 ± A . ] 4 . i . 9 l ; ' $ECH> ^$%> Q * ( *$% 2 H ^$%> ! 8$% b K *$ 9 O Fm# . " ¡ €~d} x sPvu}*{4sut }Q3?}*q | z t ́ ˆ_ L 8$% μ ' $^23 U % + Y# . 0 j# . ^>+ $% ' Y 8$=23 Fd % "„?> =@( % 2 ¶ @ @ $% j C· 8$% + Z . R 2 . # % Q 8$% E + ?@+ .$% # % 8 = $. % $ $% A# . % ECH 8$%>0 $% O$ *$% " Ji = 4 % Z$%> . 0 'CH2 c FQ/?5‹ * *$% 12 j &(@ 8$.; 9X $% $Z FI % $% \ FO$%> 6 8$% = 8 % ' o $% ƒ"•f!# 2+ . @ @ % 4 >:# . Q $% $.;h2+ . ' U@ % ' ( $% Q$% Z ( .$M .@+ % . U# . , $% 9 Z $. % 8&X $% W :Fd# H $% 9 $. % &a ̧B + $%> i# . ,/?5L$% U@ % b ( A % + $% "^S< W$%> O@+ @3 ' C? W@ % . $=$%> ]Fb C1 %coFI U$%> 0 C> (2 % \ @ @ % 4 >ƒ /1 $% $.;<Œ1 *$% ' L/1 23 $% 4 ]57 $% % 9Lh/?ŒE/?5‘ "?JW U @(; @ 8 R$%> nFb C1 %cA A$%> _ Fƒ H % $% e .> C1$%> $ #( = @ @( % 4 >e@3 .Fd % Z . 9 8l+ ' $% e23 $.$% $%>+ ^23 $%>0@ #( % O/?5X + Z@ #( % Q $% $.;h2+ . ' = * ( *$% " ?> _ % + ( H FN$%> Q@ @3 H E %94 k ' 0 HFI 'CH "Eœ $% o @( % G A # *$. $% 4 E & @ n$% ! $% $% ? #(  @ @ % 4 >ƒ" ST Wœ( $% ]»( ƒC? ( . % 23 A #( O ] + U$%> R94 $%> % 9= F $ ("œ $% -1⁄4_ . % 23 m R $ ( # m @ $ $% O F $%> $% $.;h2 . ' W@( % ' $% ' 7$%> =/?5D 9 % 8$%> ± + :$%> > (2 % @ @ % >N"GJi _@ % . $H # H &(@3 % $ + % .# 8$% H 0œ( $% U1⁄2 , &(@ RCH> R #(  * *$% „@3 .Fd % GC1 œ $% 3⁄4 "œ( ; $% 0¿-@ % @3 . H $%> 1$% R @ % 4 ! # _/?ŒE/?5j@( % ' $% " ST Aœ( $% AÀ C? H ( . # . G@ % 1 $.$% @ $% 7 $ $% 9 $% 9Q 2(; $% X W $% $' ̧ + ]l + 8 W jœ( $% jž( mC1 #+ CH 8$%>0 . _Fd# $%#( % M & $% . E$% A #( EC? %ca" 2. MOTIVATING EXAMPLE ST j$%> , . $% ƒ ?C? ^ . % 23 ^ ] j . % W X % A; E * ( *$% H + Z .> 'C\CH> Z23 $%>^@ #( % Q 23 $% 4 + $% $.;h2 . ' … $%> ( Fd _$% j@( % ( ( 09 Y % ; $% "BJW M$ c4 M$%> ( 0 F M % + $% _ Q % @( % . $ $% 4 M ' . " ST = *$E * ( *$% # . % E@ % b ( !Fd ' 2 cZ 0 8$% 1$%> $E$%> c4 ' e Z . c4 ' ƒ 1# . 9:CH> >Y@ % l+ % ^Fd % ' X$% ] ' % 23 # $A$%> = .@3 8l+ ^ $% % *$% F! >e# . '"…5 , & @ 1

Journal ArticleDOI
TL;DR: This paper presents a framework of personalization expert by combining collaborative filtering method and association rule mining technique to overcome problems that traditional personalized systems have.
Abstract: Web personalization has been providing electronic businesses with ways to keep existing customers and to obtain new ones. There are two approaches for providing personalized service: a content-based approach and a collaborative filtering approach. In the content-based approach, it is not easily applied to web objects (pages, images, sounds, etc) which are represented by multimedia data type information. Collaborative filtering approaches have cold-start problem. More serious weakness of collaborative filtering is that rating schemes can only be applied to homogenous domain information. In this paper, we present a framework of personalization expert by combining collaborative filtering method and association rule mining technique to overcome problems that traditional personalized systems have. Since multimedia data type web object cannot be easily analyzed, we adopted a collaborative filtering method that considers each object as an item, and attempts a personalized service. Similar users of each domain object are found as the result of the collaborative filtering method. These similar users’ web object access data is used by apriori algorithm to discover object association rules.

Journal ArticleDOI
TL;DR: This article presents a task-focused approach to recommendation that is entirely independent of the type of content involved and leverages robust, high-performance, commercial software.
Abstract: A technique that correlates database items to a task adds content-independent context to a recommender system based solely on user interest ratings. In this article, we present a task-focused approach to recommendation that is entirely independent of the type of content involved. The approach leverages robust, high-performance, commercial software. We have implemented it in a live movie recommendation site and validated it with empirical results from user studies.

Book ChapterDOI
05 Sep 2001
TL;DR: This work develops a recommendation system, termed Yoda, that is designed to support large-scale Web-based applications requiring highly accurate recommendations in real-time and introduces a hybrid approach that combines collaborative filtering (CF) and content-based querying to achieve higher accuracy.
Abstract: Recommendation systems are applied to personalize and customize the Web environment. We have developed a recommendation system, termed Yoda, that is designed to support large-scale Web-based applications requiring highly accurate recommendations in real-time. With Yoda, we introduce a hybrid approach that combines collaborative filtering (CF) and content-based querying to achieve higher accuracy. Yoda is structured as a tunable model that is trained off-line and employed for real-time recommendation on-line. The on-line process benefits from an optimized aggregation function with low complexity that allows realtime weighted aggregation of the soft classification of active users to predefined recommendation sets. Leveraging on localized distribution of the recommendable items, the same aggregation function is further optimized for the off-line process to reduce the time complexity of constructing the pre-defined recommendation sets of the model. To make the off-line process scalable furthermore, we also propose a filtering mechanism, FLSH, that extends the Locality Sensitive Hashing technique by incorporating a novel distance measure that satisfies specific requirements of our application. Our end-to-end experiments show while Yoda's complexity is low and remains constant as the number of users and/or items grow, its accuracy surpasses that of the basic nearest-neighbor method by a wide margin (in most cases more than 100%).

Journal ArticleDOI
TL;DR: The general architecture and function of an intelligent recommendation system aimed at supporting a leisure traveller in the task of selecting a tourist destination, bundling a set of products and composing a plan for the travel is described.
Abstract: This paper describes the general architecture and function of an intelligent recommendation system aimed at supporting a leisure traveller in the task of selecting a tourist destination, bundling a set of products and composing a plan for the travel. The system enables the user to identify his own destination and to personalize the travel by aggregating elementary items (additional locations to visit, services and activities). Case-Based Reasoning techniques enable the user to browse a repository of past travels and make possible the ranking of the elementary items included in a recommendation when these are selected from a catalogue. The system integrates data and information originating from external, already existent, tourist portals exploiting an XML-based mediator architecture, data mapping techniques, similarity-based retrieval and online analytical processing.

01 Jan 2001
TL;DR: This paper examines a method that is based on a new type of association patterns, which differently from existing approaches, considers all the specific characteristics of the Web-user navigation, and experimental results indicate its superiority over existing methods.
Abstract: The problem of predicting web-user accesses has recently attracted significant attention. Several algorithms have been proposed, which find important applications, like user profiling, recommender systems, web prefetching, design of adaptive web sites, etc. In all these applications the core issue is the developement of an effective prediction algorithm. In this paper, we focus on web-prefetching, because of its importance in reducing user perceived latency present in every Web-based application. The proposed method can be easily extended to the other aforementioned applications. Prefetching refers to the mechanism of deducing forthcoming page accesses of a client, based on access log information. We examine a method that is based on a new type of association patterns, which differently from existing approaches, considers all the specific characteristics of the Web-user navigation. Experimental results indicate its superiority over existing methods. Index Terms — Prediction, Web Log Mining, Prefetching, Association Rules, Data Mining.

Proceedings ArticleDOI
Natalie S. Glance1
01 Jan 2001
TL;DR: A new software agent, the community search assistant, which recommends related searches to users of search engines, and transforms single user usage of information networks into collaborative usage: all users can tap into the knowledge base of queries submitted by others.
Abstract: This paper describes a new software agent, the community search assistant, which recommends related searches to users of search engines. The community search assistant enables communities of users to search in a collaborative fashion. All queries submitted by the community are stored in the form of a graph. Links are made between queries that are found to be related. Users can peruse the network of related queries in an ordered way: following a path from a first cousin, to a second cousin to a third cousin, etc. to a set of search results. The first key idea behind the use of query graphs is that the determination of relatedness depends on the documents returned by the queries, not on the actual terms in the queries themselves. The second key idea is that the construction of the query graph transforms single user usage of information networks (e.g. search) into collaborative usage: all users can tap into the knowledge base of queries submitted by others.

01 Jan 2001
TL;DR: The experimental results indicate that with proper data preparation, the clustering-based approach to collaborative filtering can achieve dramatic improvements in terms of recommendation effectiveness, while maintaining the computational advantage over the direct approaches such as the k-NearestNeighbor technique.
Abstract: Recommender systems based on collaborative filtering usually require real-time comparison of users’ ratings on objects In the context of Web personalization, particularly at the early stages of a visitor’s interaction with the site (ie, before registration or authentication), recommender systems must rely on anonymous clickstream data The lack of explicit user ratings and the shear amount of data in such a setting poses serious challenges to standard collaborative filtering techniques in terms of scalability and performance Offline clustering of users transactions can be used to improve the scalability of collaborative filtering, however, this is often at the cost of reduced recommendation accuracy In this paper we study the impact of various preprocessing techniques applied to clickstream data, such as clustering, normalization, and significance filtering, on collaborative filtering Our experimental results, performed on real usage data, indicate that with proper data preparation, the clustering-based approach to collaborative filtering can achieve dramatic improvements in terms of recommendation effectiveness, while maintaining the computational advantage over the direct approaches such as the k-NearestNeighbor technique

Patent
05 Dec 2001
TL;DR: In this article, a computerized method for recommending items such as books and audio compact disks is proposed, which does not include pre-computed similarity factors measuring similarity between users, rather, when an advisee requests a recommendation, similarity measures are computed comparing the advisee to other users, and the similarity measures associated with the other users.
Abstract: Means and a computerized method for recommending items such as books and audio compact disks. For each item, a user profile includes ratings provided by users of the system. Unlike present recommendation systems, the user profiles do not include pre-computed similarity factors measuring similarity between users. Rather, when an advisee requests a recommendation, similarity measures are computed comparing the advisee to other users, and the similarity measures are associated with the other users. A subset of the users is selected, where the subset includes the users most similar to the advisee. A recommendation is made based on the ratings by the members of the selected subset.

Proceedings Article
08 Jul 2001
TL;DR: This work shows how the principle underlying privacy should be simple can be made precise in several diverse settings --- in the use of marketing surveys by a vendor designing a product; and in the design of collaborative filtering and recommendation systems, where the value of each user's participation is sought.
Abstract: As individuals increasingly take advantage of on-line services, the value of the private information they possess emerges as a problem of fundamental concern. We believe that the principle underlying privacy should be simple: Individuals are entitled to control the dissemination of private information, disclosing it as part of a transaction only when they are fairly compensated. We show how this principle can be made precise in several diverse settings --- in the use of marketing surveys by a vendor designing a product; and in the design of collaborative filtering and recommendation systems, where we seek to quantify the value of each user's participation. Our approach draws on the analysis of coalitional games, making use of the core and Shapley value of such games as fair allocation principles.

Book ChapterDOI
16 Nov 2001
TL;DR: This paper proposes to extend traditional two-dimensional user/item recommender systems to support multiple dimensions, as well as comprehensive profiling and hierarchical aggregation (OLAP) capabilities.
Abstract: In this paper, we present a new data-warehousing-based approach to recommender systems. In particular, we propose to extend traditional two-dimensional user/item recommender systems to support multiple dimensions, as well as comprehensive profiling and hierarchical aggregation (OLAP) capabilities. We also introduce a new recommendation query language RQL that can express complex recommendations taking into account the proposed extensions. We describe how these extensions are integrated into a framework that facilitates more flexible and comprehensive user interactions with recommender systems.

Journal ArticleDOI
TL;DR: A simple analytical framework for recommendation systems is introduced, including a basis for defining the utility of such a system, and probabilistic analyses of algorithms within this framework yield insights into how much utility can be derived from knowledge of past user actions.

Proceedings ArticleDOI
01 Jan 2001
TL;DR: A collaborative filtering algorithm which automatically rates reviewers, and incorporates the quality of the reviewer into the metric of merit for the paper, which seems to provide all the benefits of the current peer review system, while at the same time being much more flexible.
Abstract: The current system for scholarly information dissemination may be amen able to significant improvement. In particular, going from the current system of journal publication to one of self-distributed documents offers significant cost and timeliness advantages. A major concern with such alternatives is how to provide the value currently afforded by the peer review system.Here we propose a mechanism that could plausibly supply such value. In the peer review system, papers are judged meritorious if good reviewers give them good reviews. In its place, we propose a collaborative filtering algorithm which automatically rates reviewers, and incorporates the quality of the reviewer into the metric of merit for the paper. Such a system seems to provide all the benefits of the current peer review system, while at the same time being much more flexible.We have implemented a number of parameterized variations of this algorithm, and tested them on data available from a quite different application. Our initial experiments suggest that the algorithm is in fact ranking reviewers reasonably.