Wyniki wyszukiwania - BazTech

1

Stereotype-Aaware Collaborative Filtering

Frisch Gabriel, Leger Jean-Benoist, Grandvalet Yves

Annals of Computer Science and Information Systems

|

2021

|

Vol. 25

69--79

EN

In collaborative filtering, recommendations are made using user feedback on a few products. In this paper, we show that even if sensitive attributes are not used to fit the models, a disparate impact may nevertheless affect recommendations. We propose a definition of fairness for the recommender system that expresses that the ranking of items should be independent of sensitive attribute. We design a co-clustering of users and items that processes exogenous sensitive attributes to remove their influence to return fair recommendations. We prove that our model ensures approximately fair recommendations provided that the classification of users approximately respects statistical parity.

2

Pix2Trips - a System Supporting Small Groups of Urban Tourists

Kwaśnicka Halina, Ovedenski Tsvetan

Annals of Computer Science and Information Systems

|

2021

|

Vol. 25

141--145

EN

Group recommendation systems are the subject of many publications, but still is a gap between research results and group decision support systems' needs. Tourists often do not know which attractions they would like to visit, our Pix2Trips system asks the group's members to indicate images that, in their opinion, they would like. Pix2Trips models the group's preferences and adjusts it to the proposed places' models. Some tourist places in Wroclaw city, Poland, were used in experiments. The paper presents the system's components and discusses the results of the experiments. Conclusions indicate the good overall evaluation of the Pix2Trips system.

3

Dynamic Clustering Personalization for Recommending Long Tail Items

Vinícius de Sousa Silva Diogo, Araújo Durão Frederico

Annals of Computer Science and Information Systems

|

2020

|

Vol. 21

417--425

EN

Recommendation strategies are used in several contexts in order to bring potential users closer to products with a strong probability of interest. When recomendations focus on niche items, they are called recommendations in the long tail. In these cases, they also look for less popular items and try to find your target custumer, niche market. This paper proposes a long tail recommendation approach that prioritizes relevance, diversity and popularity of recommended items. For that, a hybrid approach based on two techniques are used. The first is clustering with dynamic parameters that adapt from according to the dataset used and the second is a type of Markov chains for to calculate the distance of interest of a user to an item of relevance for this user. The results show that the techniques used have a better relevance indexes at the same time more diverse and less popular recommendations.

4

Rozmyta mapa kognitywna jako inteligentny system rekomendacyjny zasobów strony internetowej

Jastriebow A., Kubuś Ł., Poczęta K.

Informatyka, Automatyka, Pomiary w Gospodarce i Ochronie Środowiska

|

2017

|

T. 7, nr 4

74--78

PL

Artykuł poświęcony jest budowie i analizie inteligentnego systemu rekomendacyjnego zasobów bazującego na rozmytej mapie kognitywnej. Opracowany system pozwala wskazać zasoby strony internetowej, którymi może być zainteresowany potencjalny użytkownik. Zasoby te są określane na podstawie aktywności innych użytkowników serwisu. Bazując na zbiorze anonimowo zebranych danych historycznych opracowano rozmytą mapę kognitywną, której czynniki odpowiadają poszczególnym zasobom strony internetowej. Wagi powiązań między nimi określono na podstawie liczby użytkowników odwiedzających poszczególne zasoby.

EN

This paper is devoted to the construction and analysis of the intelligent recommendation system for website resources based on fuzzy cognitive map. The developed system allows to identify resources, which may be interested in a potential user. These resources are determined on the basis of website users activity. Fuzzy cognitive map was develop using the dataset with anonymous collected historical data. The concepts of fuzzy cognitive map are identifiers of resources of website. Weights of the connection between them have been established based on the number of users visiting the resources.

5

Adaptive test selection for factorization-based surrogate fitness in genetic programming

Krawiec K., Liskowski P.

Foundations of Computing and Decision Sciences

|

2017

|

Vol. 42, No. 4

339--358

EN

Genetic programming (GP) is a variant of evolutionary algorithm where the entities undergoing simulated evolution are computer programs. A fitness function in GP is usually based on a set of tests, each of which defines the desired output a correct program should return for an exemplary input. The outcomes of interactions between programs and tests in GP can be represented as an interaction matrix, with rows corresponding to programs in the current population and columns corresponding to tests. In previous work, we proposed SFIMX, a method that performs only a fraction of interactions and employs non-negative matrix factorization to estimate the outcomes of remaining ones, shortening GP’s runtime. In this paper, we build upon that work and propose three extensions of SFIMX, in which the subset of tests drawn to perform interactions is selected with respect to test difficulty. The conducted experiment indicates that the proposed extensions surpass the original SFIMX on a suite of discrete GP benchmarks.

6

E-tourism recommender systems: a survey and development perspectives

Artemenko O., Kunanets O., Pasichnyk V.

ECONTECHMOD : An International Quarterly Journal on Economics of Technology and Modelling Processes

|

2017

|

Vol. 6, No 2

91--95

EN

This paper describes main modern tendencies of the design and development of intelligent information technologies, implementing process of generating recommendations in tourism recommender systems. The basic trends of the e-tourism information technology tools are analyzed to show importance of creation for multitask, multykontent mobile e-tourism recommender systems with decision support functions in terms of the tourism group, which is regarded as a single collective user.

7

Latency of Neighborhood Based Recommender Systems

Chojnacki S., Kłopotek M.

Fundamenta Informaticae

|

2015

|

Vol. 139, nr 3

229--248

EN

Latency of user-based and item-based recommenders is evaluated. The two algorithms can deliver high quality predictions in dynamically changing environments. However, their response time depends not only on the size, but also on the structure of underlying datasets. This constitutes a major drawback when compared to two other competitive approaches i.e. content-based and modelbased systems. Therefore, we believe that there exists a need for comprehensive evaluation of the latency of the two algorithms. During a typical worst case scenario analysis of collaborative filtering algorithms two assumption are made. The first assumption says that data are stored in dense collections. The second assumption states that large amount of computations can be performed in advance during the training phase. As a result it is advised to deploy user-based system when the number of users is relatively small. Item-based algorithms are believed to have better technical properties when the number of items is small. We consider a situation in which the two assumptions are not necessarily met. We show that even though the latency of the two methods depends heavily on the proportion of users to items, this factor does not differentiate the two methods. We evaluate the algorithms with several real-life datasets. We augment the analysis with both graph-theoretical and experimental techniques.

8

Utilizing User Access Sequence to Mitigate the Cold Start Problem in Collaborative Filtering Recommendation

Deng X.

Przegląd Elektrotechniczny

|

2013

|

R. 89, nr 1b

217--220

EN

Collaborative filtering (CF) is one of the most successful recommending techniques, but it suffers from the cold start problem which severely affected the quality of recommendation. To address this problem, we propose a novel hybrid approach, named UAS-CF, which incorporates user access sequence into traditional CF for improving the quality of recommendation. Experiments on three datasets were carried out to evaluate the performance of our method. Our results show that our approach outperforms other methods and improves recommendation quality effectively.

PL

W artykule zaprezentowano nowe podejście UAS-CF do obsługi poleceń, które włącza sekwencję dostępu użytkownika do klasycznego filtrowania uwspólnionego (ang. Collaborative Filtering), w celu polepszenia jakości rekomendacji. Badania eksperymentalne, przeprowadzone na trzech sekwencjach danych, wykazują wysoką jakość rekomendacji w porównaniu z innymi metodami.

9

Collaborative filtering recommender systems in music recommendation

Kużelewska U., Ducki R.

Advances in Computer Science Research

|

2013

|

Nr 10

67--79

EN

Nowadays, the primary place of information exchange is the internet. Its features, such as: availability, unlimited capacity and diversity of information influenced its unrivalled popularity, making the internet a powerful platform for storage, dissemination and retrieval of information. On the other hand, the internet data are highly dynamic and unstructured. As a result, the internet users face the problem of data overload. Recommender systems help the users to find the products, services or information they are looking for. The article presents a recommender system for music artist recommendation. It is composed of user-based as well as item-based procedures, which can be selected dynamically during a user’s session. This also includes different similarity measures. The following measures are used to assess the recommendations and adapt the appropriate procedure: RMSE, MAE, Precision and Recall. Finally, the generated recommendations and calculated similarities among artists are compared with the results from LastFM service.

PL

W obecnych czasach głównym miejscem wymiany informacji jest internet. Jego cechy, takie jak: wysoka dostępność, nieograniczona pojemność i różnorodność informacji wpłynęły na jego niezrównana popularność. W ten sposób internet stał się potężną platformą do przechowywania, rozpowszechniania i udostępniania informacji. Z drugiej strony, dane internetowe są bardzo dynamiczne i niestrukturalizowane. W rezultacie, użytkownicy internetu muszą radzić sobie z problemem przeładowania danych. Systemy rekomendujące służą pomocą użytkownikom w celu znalezienia poszukiwanych produktów, usług lub informacji. W artykule przedstawiono system rekomendujący artystów muzycznych. Składa się on z procedur typu user-based oraz item-based oraz różnych sposobów szacowania podobieństwa, które mogą˛ być zmieniane dynamicznie podczas sesji użytkownika. Do oceny list rekomendacji wykorzystano następujące miary: RMSE, MAE, Precision i Recall. Dodatkowo, wygenerowane rekomendacje i obliczone podobieństwa miedzy artystami są porównywane z wynikami z serwisu LastFM .

10

Random graphs for performance evaluation of recommender systems

Chojnacki Sz., Kłopotek M. A.

Control and Cybernetics

|

2011

|

Vol. 40, no 2

237-257

EN

The purpose of this article is to introduce a new analytical framework dedicated to measuring performance of recommender systems. A standard approach is to assess the quality of a system by means of accuracy related statistics. However, the specificity of the environments in which recommender systems are deployed requires paying much attention to speed and memory requirements of the algorithms. Unfortunately, it is implausible to assess accurately the complexity of various algorithms with formal tools. This can be attributed to the fact that such analyses are usually based on an assumption of dense representation of underlying data structures. In real life, though, the algorithms operate on sparse data and are implemented with collections dedicated for them. Therefore, we propose to measure the complexity of recommender systems with artificial datasets that posses real-life properties. We utilize a recently developed bipartite graph generator to evaluate how the state-of-art recommender system behavior is determined and diversified by topological properties of the generated datasets.

11

Random graph generator for bipartite networks modeling

Chojnacki Sz., Kłopotek M. A.

Control and Cybernetics

|

2011

|

Vol. 40, no 3

697-714

EN

The purpose of this article is to introduce a new bipartite graph generation algorithm. Bipartite graphs consist of two types of nodes and edges join only nodes of different types. This data structure appears in various applications (e.g. recommender systems or text clustering). Both real-life datasets and formal tools enable us to evaluate only a limited set of properties of the algorithms that are used in such situations. Therefore, artificial datasets are needed to enhance development and testing of the algorithms. Our generator can be used to produce a wide range of synthetic datasets.

12

An improved recommender system to avoid the persistent information overload in a university digital library

Porcel C., Morales del Castillo J. M., Cobo M. J., Ruiz A. A., Herrera-Viedma E.

Control and Cybernetics

|

2010

|

Vol. 39, no 4

899-923

EN

Nowadays we are continuously bombarded with a lot of information, and because of it we have serious problems with accessing the relevant information, that is, we suffer from the information overload problems. Recommender systems have been applied successfully to avoid the information overload in different domains, but the number of electronic resources daily generated keeps growing and the problem rises again. Therefore, we find a persistent problem of information overload. In this paper we propose an improved recommender system to avoid the persistent information overload found in a University Digital Library. The idea is to include a memory to remember selected resources but not recommended to the user, and in such a way, the system could incorporate them in future recommendations to complete the set of filtered resources, for example, if there are a few resources to be recommended or if the user wishes output obtained by combination of resources selected in different recommendation rounds.

13

Pamięciowe algorytmy kooperatywnej filtracji

Cichosz P.

Elektronika : konstrukcje, technologie, zastosowania

|

2009

|

Vol. 50, nr 8

267-271

PL

Zadanie kooperatywnej filtracji, dla pewnego zbioru użytkowników i obiektów, polega na prognozowaniu ocen, jakie przyznaliby użytkownicy nieznanym sobie obiektom, na podstawie znanych ocen przyznanych przez tych użytkowników innym obiektom oraz przez innych użytkowników tym samym obiektom. Rozwiązanie tego zadania umożliwia tworzenie inteligentnych systemów rekomendujących (np. dla elektronicznego handlu) charakteryzujących się wysokim stopniem trafności i indywidualizacji ofert. Artykuł jest poświęcony algorytmom kooperatywnej filtracji opartym na zapamiętywaniu znanych ocen i wykorzystywaniu ich do identyfikacji "najbliższych sąsiadów", na podstawie których wyznaczana jest prognoza. Przedstawiony jest wspólny schemat tego typu algorytmów, jego najbardziej typowa konkretyzacja oraz pozostające do zbadania możliwości poprawy dokładności lub efektywności.

EN

The collaborative filtering task, for a given set of users and items, consists in predicting ratings that the users would assign to unknown items based on known ratings assigned by the same users to similar items or by similar users to the same items. Solutions to this task enable creating intelligent recommender systems (e.g., for e-commerce). The paper is devoted to collaborative filtering algorithms based on storing known ratings and using them to identify the nearest neighbors used to calculate the predictions. A common basic template of such algorithms is presented with its most typical instantiation and remaining possibilities of improving accuracy or efficiency.

14

Investigation into the application of graph neural networks to large-scale recommender systems

Pucci A., Gori M., Hagenbuchner M., Scarselli F., Tsoi A. C.

Systems Science

|

2006

|

Vol. 32, no 4

17-26

EN

Recent developments with Neural Networks produced models which are capable of encoding graph structured data. The most promising and arguably the most capable of these methods are the Graph Neural Networks (GNNs). This paper considers the GNN for the application to recommender system learning problem. It will be shown that the GNN can generally exploit relational information in such a graph structured domain but experiments also revealed some interesting limitations of the GNN. Experiments were conducted on a relatively large set o real world data from MovieLens. The dataset that has been widely used as a benchmark problem. The careful analysis of these data led to the discovery of some intriguing properties which helped to explain the problems and limitations of the GNN when dealing with this learning problem.