In this thesis, various machine learning domains have been combined in order to build a video recommender system that is based on object detection. The work combines two extensively studied research fields, recommender systems and computer vision, that also are rapidly growing and popular techniques on commercial markets. To investigate the performance of the approach, three different content-based recommender systems have been implemented at Spotify, which are based on the following video features: object detections, titles and descriptions, and user preferences. These systems have then been evaluated and compared against each other together with their hybridized result. Two algorithms have been implemented, the prediction and the top-N algorithm, where the former is the more reliable source for evaluating the system's performance. The evaluation of the system shows that the overall performance scores for predicting values of the users' liked and disliked videos are in the range from about 40 % to 70 % for the prediction algorithm and from about 15 % to 70 % for the top-N algorithm. The approach based on object detection performs worse in comparison to the other approaches. Hence, there seems to be is a low correlation between the user preferences and the video contents in terms of object detection data. Therefore, this data is not very suitable for describing the content of videos and using it in the recommender system. However, the results of this study cannot be generalized to apply for other systems before the approach has been evaluated in other environments and for various data sets. Moreover, there are plenty of room for refinements and improvements to the system, as well as there are many interesting research areas for future work.

Conselho Nacional de Desenvolvimento Científico e Tecnológico / Hoje em dia, a quantidade de usuários que utilizam sites de comércio eletrônico para realizar compras está aumentando muito, principalmente devido à facilidade e rapidez. Muitos sites de comércio eletrônico, diferentemente das lojas físicas, disponibilizam aos seus usuários uma grande variedade de produtos e serviços, e os usuários podem ter muita dificuldade em encontrar produtos de sua preferência. Normalmente, a preferência por um produto pode ser influenciada pela aparência visual da imagem do produto. Neste contexto, os Sistemas de Recomendação de produtos que estão associados a Imagens (SRI) tornaram-se indispensáveis para ajudar os usuários a encontrar produtos que podem ser, eventualmente, agradáveis ou úteis para eles. Geralmente, os SRI usam o comportamento passado dos usuários (cliques, compras, críticas, avaliações, etc.) e/ou atributos de produtos para definirem as preferências dos usuários. Um dos principais desafios enfrentados em SRI é a necessidade de o usuário fornecer algumas informações sobre suas preferências sobre os produtos, a fim de obter novas recomendações do sistema. Infelizmente, os usuários nem sempre estão dispostos a fornecer tais informações de forma explícita. Assim, a fim de lidar com esse desafio, os métodos para obtenção de informações de forma implícita do usuário são desejáveis. Neste trabalho, propõe-se investigar em que medida informações sobre atenção visual do usuário podem ajudar a melhorar a predição de avaliação e consequentemente produzir SRI mais precisos. É também objetivo deste trabalho o desenvolvimento de dois novos métodos, um método baseado em Filtragem Colaborativa (FC) que combina avaliações e dados de atenção visual para representar o comportamento passado dos usuários, e outro método baseado no conteúdo dos itens, que combina atributos textuais, características visuais e dados de atenção visual para compor o perfil dos itens. Os métodos propostos foram avaliados em uma base de imagens de pinturas e uma base de imagens de roupas. Os resultados experimentais mostram que os métodos propostos neste trabalho possuem ganhos significativos em predição de avaliação e precisão na recomendação quando comparados ao estado-da-arte. Vale ressaltar que as técnicas propostas são flexíveis, podendo ser utilizadas em outros cenários que exploram a atenção visual dos itens recomendados. / Nowadays, the amount of users using e-commerce sites for shopping is greatly increasing, mainly due to the easiness and rapidity of this way of consumption. Many e-commerce sites, differently from physical stores, can offer their users a wide range of products and services, and the users can find it very difficult to find products of your preference. Typically, your preference for a product can be influenced by the visual appearance of the product image. In this context, Image Recommendation Systems (IRS) have become indispensable to help users to find products that may possibly pleasant or be useful to them. Generally, IRS use past behavior of users (clicks, purchases, reviews, ratings, etc.) and/or attributes of the products to define the preferences of users. One of the main challenges faced by IRS is the need of the user to provide some information about his / her preferences on products in order to get further recommendations from the system. Unfortunately, users are not always willing to provide such information explicitly. So, in order to cope with this challenge, methods for obtaining user’s implicit feedback are desirable. In this work, the author propose an investigation to discover to which extent information concerning user visual attention can help improve the rating prediction hence produce more accurate IRS. This work proposes to develop two new methods, a method based on Collaborative Filtering (CF) which combines ratings and data visual attention to represent the past behavior of users, and another method based on the content of the items, which combines textual attributes, visual features and visual attention data to compose the profile of the items. The proposed methods were evaluated in a painting dataset and a clothing dataset. The experimental results show significant improvements in rating prediction and precision in recommendation when compared to the state-of-the-art. It is worth mentioning that the proposed techniques are flexible and can be applied in other scenarios that exploits the visual attention of the recommended items. / Tese (Doutorado)

No description available.

This thesis deals with the field of recommendation systems using deep neural networks and their use in book recommendation. There are the main traditional recommender systems analysed and their representations are summarized, as well as systems with more advanced techniques based on machine learning. The core of the thesis is to use convolutional neural networks for natural language processing and create a hybrid book recommendation system. Suggested system includes matrix factorization and make recommendation based on user ratings and book metadata, including texts descriptions. I designed two models, one with bag-of-words technique and one with convolutional neural network. Both of them defeat baseline methods. On the created data set, that was created from the Goodreads, model with CNN beats model with BOW.

This thesis deals with the field of Recommendation systems using Deep Neural Networks and their use in book recommendation. There are the main traditional recommender systems analysed and their representations are summarized, as well as systems with more advancec techniques based on machine learning.. The core of the thesis is the use of convolutional neural networks for natural language processing and the creation of a book recommendation system. Suggested system make recommendation based on user data, including user reviews and book data, including full texts.

Depuis ses fondements, le domaine de l’Interaction Homme-Machine (IHM) est marqué par le souci constant de concevoir et de produire des systèmes numériques utiles et utilisables, c’est-à-dire adaptés aux utilisateurs dans leur contexte. Vu le développement exponentiel des recherches dans les IHM, deux états des lieux s’imposent dans les environnements en ligne : le concept de confiance et le comportement de l’usager. Ces deux états ne cessent de proliférer dans la plupart des solutions conçues et sont à la croisée des travaux dans les interfaces de paiements en ligne et dans les systèmes de recommandation. Devant les progrès des solutions conçues, l’objectif de cette recherche réside dans le fait de mieux comprendre les différents enjeux dans ces deux domaines, apporter des améliorations et proposer de nouvelles solutions adéquates aux usagers en matière de perception et de comportement en ligne. Outre l’état de l’art et les problématiques, ce travail est divisé en cinq parties principales, chacune contribue à mieux enrichir l’expérience de l’usager en ligne en matière de paiement et recommandations en ligne : • Analyse des multi-craintes en ligne : nous analysons les différents facteurs des sites de commerce électronique qui influent directement sur le comportement des consommateurs en matière de prise de décision et de craintes en ligne. Nous élaborons une méthodologie pour mesurer avec précision le moment où surviennent la question de la confidentialité, les perceptions en ligne et les craintes de divulgation et de pertes financières. • Intégration de personnalisation, contrôle et paiement conditionnel : nous proposons une nouvelle plateforme de paiement en ligne qui supporte à la fois la personnalisation et les paiements multiples et conditionnels, tout en préservant la vie privée du détenteur de carte. • Exploration de l’interaction des usagers en ligne versus la sensibilisation à la cybersécurité : nous relatons une expérience de magasinage en ligne qui met en relief la perception du risque de cybercriminalité dans les activités en ligne et le comportement des utilisateurs lié à leur préoccupation en matière de confidentialité. • Équilibre entre utilité des données et vie privée : nous proposons un modèle de préservation de vie privée basé sur l’algorithme « k-means » et sur le modèle « k-coRating » afin de soutenir l’utilité des données dans les recommandations en ligne tout en préservant la vie privée des usagers. • Métrique de stabilité des préférences des utilisateurs : nous ciblons une meilleure méthode de recommandation qui respecte le changement des préférences des usagers par l’intermédiaire d’un réseau neural. Ce qui constitue une amélioration à la fois efficace et performante pour les systèmes de recommandation. Cette thèse porte essentiellement sur quatre aspects majeurs liés : 1) aux plateformes des paiements en ligne, 2) au comportement de l’usager dans les transactions de paiement en ligne (prise de décision, multi-craintes, cybersécurité, perception du risque), 3) à la stabilité de ses préférences dans les recommandations en ligne, 4) à l’équilibre entre vie privée et utilité des données en ligne pour les systèmes de recommandation. / Technologies in Human-Machine Interaction (HMI) are playing a vital role across the entire production process to design and deliver advanced digital systems. Given the exponential development of research in this field, two concepts are largely addressed to increase performance and efficiency of online environments: trust and user behavior. These two extents continue to proliferate in most designed solutions and are increasingly enriched by continuous investments in online payments and recommender systems. Along with the trend of digitalization, the objective of this research is to gain a better understanding of the various challenges in these two areas, make improvements and propose solutions more convenient to the users in terms of online perception and user behavior. In addition to the state of the art and challenges, this work is divided into five main parts, each one contributes to better enrich the online user experience in both online payments and system recommendations: • Online customer fears: We analyze different components of the website that may affect customer behavior in decision-making and online fears. We focus on customer perceptions regarding privacy violations and financial loss. We examine the influence on trust and payment security perception as well as their joint effect on three fundamentally important customers’ aspects: confidentiality, privacy concerns and financial fear perception. • Personalization, control and conditional payment: we propose a new online payment platform that supports both personalization and conditional multi-payments, while preserving the privacy of the cardholder. • Exploring user behavior and cybersecurity knowledge: we design a new website to conduct an experimental study in online shopping. The results highlight the impact of user’s perception in cybersecurity and privacy concerns on his online behavior when dealing with shopping activities. • Balance between data utility and user privacy: we propose a privacy-preserving method based on the “k-means” algorithm and the “k-coRating” model to support the utility of data in online recommendations while preserving user’s privacy. • User interest constancy metric: we propose a neural network to predict the user’s interests in recommender systems. Our aim is to provide an efficient method that respects the constancy and variations in user preferences. In this thesis, we focus on four major contributions related to: 1) online payment platforms, 2) user behavior in online payments regarding decision making, multi-fears and cyber security 3) user interest constancy in online recommendations, 4) balance between privacy and utility of online data in recommender systems.

Help documents are important in an organization to use the technology applications licensed from a vendor. Customers and internal employees frequently use and interact with the help documents section to use the applications and know about the new features and developments in them. Help documents consist of various knowledge base materials, question and answer documents and help content. In day- to-day life, customers go through these documents to set up, install or use the product. Recommending similar documents to the customers can increase customer engagement in the product and can also help them proceed without any hurdles. The main aim of this study is to build a recommendation system by exploring different machine-learning techniques to recommend the most relevant and similar help document to the user. To achieve this, in this study a hybrid-based recommendation system for help documents is proposed where the documents are recommended based on similarity of the content using content-based filtering and similarity between the users using collaborative filtering. Finally, the recommendations from content-based filtering and collaborative filtering are combined and ranked to form a comprehensive list of recommendations. The proposed approach is evaluated by the internal employees of the company and by external users. Our experimental results demonstrate that the proposed approach is feasible and provides an effective way to recommend help documents.

Les étudiants gradués et les professeurs (les chercheurs, en général), accèdent, passent en revue et utilisent régulièrement un grand nombre d’articles, cependant aucun des outils et solutions existants ne fournit la vaste gamme de fonctionnalités exigées pour gérer correctement ces ressources. En effet, les systèmes de gestion de bibliographie gèrent les références et les citations, mais ne parviennent pas à aider les chercheurs à manipuler et à localiser des ressources. D'autre part, les systèmes de recommandation d’articles de recherche et les moteurs de recherche spécialisés aident les chercheurs à localiser de nouvelles ressources, mais là encore échouent dans l’aide à les gérer. Finalement, les systèmes de gestion de contenu d'entreprise offrent les fonctionnalités de gestion de documents et des connaissances, mais ne sont pas conçus pour les articles de recherche. Dans ce mémoire, nous présentons une nouvelle classe de systèmes de gestion : système de gestion et de recommandation d’articles de recherche. Papyres (Naak, Hage, & Aïmeur, 2008, 2009) est un prototype qui l’illustre. Il combine des fonctionnalités de bibliographie avec des techniques de recommandation d’articles et des outils de gestion de contenu, afin de fournir un ensemble de fonctionnalités pour localiser les articles de recherche, manipuler et maintenir les bibliographies. De plus, il permet de gérer et partager les connaissances relatives à la littérature. La technique de recommandation utilisée dans Papyres est originale. Sa particularité réside dans l'aspect multicritère introduit dans le processus de filtrage collaboratif, permettant ainsi aux chercheurs d'indiquer leur intérêt pour des parties spécifiques des articles. De plus, nous proposons de tester et de comparer plusieurs approches afin de déterminer le voisinage dans le processus de Filtrage Collaboratif Multicritère, de telle sorte à accroître la précision de la recommandation. Enfin, nous ferons un rapport global sur la mise en œuvre et la validation de Papyres. / Graduate students and professors (researchers, in general) regularly access, review, and use large amounts of research papers, yet none of the existing tools and solutions provides the wide range of functionalities required to properly manage these resources. Indeed, bibliography management systems manage the references and citations but fail to help researchers in handling and locating resources. On the other hand, research paper recommendation systems and specialized search engines help researchers to locate new resources, but again fail to help researchers in managing the resources. Finally, Enterprise Content Management systems offer the required functionalities to manage resources and knowledge, but are not designed for research literature. Consequently, we suggest a new class of management systems: Research Paper Management and Recommendation System. Through our system Papyres (Naak, Hage, & Aïmeur, 2008, 2009) we illustrate our approach, which combines bibliography functionalities along with recommendation techniques and content management tools, in order to provide a set of functionalities to locate research papers, handle and maintain the bibliographies, and to manage and share knowledge related to the research literature. Additionally, we propose a novel research paper recommendation technique, used within Papyres. Its uniqueness lies in the multicriteria aspect introduced in the process of collaborative filtering, allowing researchers to indicate their interest in specific parts of articles. Moreover, we suggest test and compare several approaches to determine the neighbourhood in the Multicriteria Collaborative Filtering process, such as to increase the accuracy of the recommendation. Finally, we report on the implementation and validation of Papyres.

Sistemas de Recomendação (SR) são softwares capazes de sugerir itens aos usuários com base no histórico de interações de usuários ou por meio de métricas de similaridade que podem ser comparadas por item, usuário ou ambos. Existem diferentes tipos de SR e dentre os que despertam maior interesse deste trabalho estão: SR baseados em conteúdo; SR baseados em conhecimento; e SR baseado em filtro colaborativo. Alcançar resultados adequados às expectativas dos usuários não é uma meta simples devido à subjetividade inerente ao comportamento humano, para isso, SR precisam de soluções eficientes e eficazes para: modelagem dos dados que suportarão a recomendação; recuperação da informação que descrevem os dados; combinação dessas informações dentro de métricas de similaridade, popularidade ou adequabilidade; criação de modelos descritivos dos itens sob recomendação; e evolução da inteligência do sistema de forma que ele seja capaz de aprender a partir da interação com o usuário. A tomada de decisão por um sistema de recomendação é uma tarefa complexa que pode ser implementada a partir da visão de áreas como inteligência artificial e mineração de dados. Dentro da área de inteligência artificial há estudos referentes ao método de raciocínio baseado em casos e da recomendação baseada em casos. No que diz respeito à área de mineração de dados, os SR podem ser construídos a partir de modelos descritivos e realizar tratamento de dados textuais, constituindo formas de criar elementos para compor uma recomendação. Uma forma de minimizar os pontos fracos de uma abordagem, é a adoção de aspectos baseados em uma abordagem híbrida, que neste trabalho considera-se: tirar proveito dos diferentes tipos de SR; usar técnicas de resolução de problemas; e combinar recursos provenientes das diferentes fontes para compor uma métrica unificada a ser usada para ranquear a recomendação por relevância. Dentre as áreas de aplicação dos SR, destaca-se a recomendação de notícias, sendo utilizada por um público heterogêneo, amplo e exigente por relevância. Neste contexto, a presente pesquisa apresenta uma abordagem híbrida para recomendação de notícias construída por meio de uma arquitetura implementada para provar os conceitos de um sistema de recomendação. Esta arquitetura foi validada por meio da utilização de um corpus de notícias e pela realização de um experimento online. Por meio do experimento foi possível observar a capacidade da arquitetura em relação aos requisitos de um sistema de recomendação de notícias e também confirmar a hipótese no que se refere à privilegiar recomendações com base em similaridade, popularidade, diversidade, novidade e serendipidade. Foi observado também uma evolução nos indicadores de leitura, curtida, aceite e serendipidade conforme o sistema foi acumulando histórico de preferências e soluções. Por meio da análise da métrica unificada para ranqueamento foi possível confirmar sua eficácia ao verificar que as notícias melhores colocadas no ranqueamento foram as mais aceitas pelos usuários / Recommendation Systems (RS) are software capable of suggesting items to users based on the history of user interactions or by similarity metrics that can be compared by item, user, or both. There are different types of RS and those which most interest in this work are content-based, knowledge-based and collaborative filtering. Achieving adequate results to user\'s expectations is a hard goal due to the inherent subjectivity of human behavior, thus, the RS need efficient and effective solutions to: modeling the data that will support the recommendation; the information retrieval that describes the data; combining this information within similarity, popularity or suitability metrics; creation of descriptive models of the items under recommendation; and evolution of the systems intelligence to learn from the user\'s interaction. Decision-making by a RS is a complex task that can be implemented according to the view of fields such as artificial intelligence and data mining. In the artificial intelligence field there are studies concerning the method of case-based reasoning that works with the principle that if something worked in the past, it may work again in a new similar situation the one in the past. The case-based recommendation works with structured items, represented by a set of attributes and their respective values (within a ``case\'\' model), providing known and adapted solutions. Data mining area can build descriptive models to RS and also handle, manipulate and analyze textual data, constituting one option to create elements to compose a recommendation. One way to minimize the weaknesses of an approach is to adopt aspects based on a hybrid solution, which in this work considers: taking advantage of the different types of RS; using problem-solving techniques; and combining resources from different sources to compose a unified metric to be used to rank the recommendation by relevance. Among the RS application areas, news recommendation stands out, being used by a heterogeneous public, ample and demanding by relevance. In this context, the this work shows a hybrid approach to news recommendations built through a architecture implemented to prove the concepts of a recommendation system. This architecture has been validated by using a news corpus and by performing an online experiment. Through the experiment it was possible to observe the architecture capacity related to the requirements of a news recommendation system and architecture also related to privilege recommendations based on similarity, popularity, diversity, novelty and serendipity. It was also observed an evolution in the indicators of reading, likes, acceptance and serendipity as the system accumulated a history of preferences and solutions. Through the analysis of the unified metric for ranking, it was possible to confirm its efficacy when verifying that the best classified news in the ranking was the most accepted by the users

Sistemas de Recomendação (SR) são softwares capazes de sugerir itens aos usuários com base no histórico de interações de usuários ou por meio de métricas de similaridade que podem ser comparadas por item, usuário ou ambos. Existem diferentes tipos de SR e dentre os que despertam maior interesse deste trabalho estão: SR baseados em conteúdo; SR baseados em conhecimento; e SR baseado em filtro colaborativo. Alcançar resultados adequados às expectativas dos usuários não é uma meta simples devido à subjetividade inerente ao comportamento humano, para isso, SR precisam de soluções eficientes e eficazes para: modelagem dos dados que suportarão a recomendação; recuperação da informação que descrevem os dados; combinação dessas informações dentro de métricas de similaridade, popularidade ou adequabilidade; criação de modelos descritivos dos itens sob recomendação; e evolução da inteligência do sistema de forma que ele seja capaz de aprender a partir da interação com o usuário. A tomada de decisão por um sistema de recomendação é uma tarefa complexa que pode ser implementada a partir da visão de áreas como inteligência artificial e mineração de dados. Dentro da área de inteligência artificial há estudos referentes ao método de raciocínio baseado em casos e da recomendação baseada em casos. No que diz respeito à área de mineração de dados, os SR podem ser construídos a partir de modelos descritivos e realizar tratamento de dados textuais, constituindo formas de criar elementos para compor uma recomendação. Uma forma de minimizar os pontos fracos de uma abordagem, é a adoção de aspectos baseados em uma abordagem híbrida, que neste trabalho considera-se: tirar proveito dos diferentes tipos de SR; usar técnicas de resolução de problemas; e combinar recursos provenientes das diferentes fontes para compor uma métrica unificada a ser usada para ranquear a recomendação por relevância. Dentre as áreas de aplicação dos SR, destaca-se a recomendação de notícias, sendo utilizada por um público heterogêneo, amplo e exigente por relevância. Neste contexto, a presente pesquisa apresenta uma abordagem híbrida para recomendação de notícias construída por meio de uma arquitetura implementada para provar os conceitos de um sistema de recomendação. Esta arquitetura foi validada por meio da utilização de um corpus de notícias e pela realização de um experimento online. Por meio do experimento foi possível observar a capacidade da arquitetura em relação aos requisitos de um sistema de recomendação de notícias e também confirmar a hipótese no que se refere à privilegiar recomendações com base em similaridade, popularidade, diversidade, novidade e serendipidade. Foi observado também uma evolução nos indicadores de leitura, curtida, aceite e serendipidade conforme o sistema foi acumulando histórico de preferências e soluções. Por meio da análise da métrica unificada para ranqueamento foi possível confirmar sua eficácia ao verificar que as notícias melhores colocadas no ranqueamento foram as mais aceitas pelos usuários / Recommendation Systems (RS) are software capable of suggesting items to users based on the history of user interactions or by similarity metrics that can be compared by item, user, or both. There are different types of RS and those which most interest in this work are content-based, knowledge-based and collaborative filtering. Achieving adequate results to user\'s expectations is a hard goal due to the inherent subjectivity of human behavior, thus, the RS need efficient and effective solutions to: modeling the data that will support the recommendation; the information retrieval that describes the data; combining this information within similarity, popularity or suitability metrics; creation of descriptive models of the items under recommendation; and evolution of the systems intelligence to learn from the user\'s interaction. Decision-making by a RS is a complex task that can be implemented according to the view of fields such as artificial intelligence and data mining. In the artificial intelligence field there are studies concerning the method of case-based reasoning that works with the principle that if something worked in the past, it may work again in a new similar situation the one in the past. The case-based recommendation works with structured items, represented by a set of attributes and their respective values (within a ``case\'\' model), providing known and adapted solutions. Data mining area can build descriptive models to RS and also handle, manipulate and analyze textual data, constituting one option to create elements to compose a recommendation. One way to minimize the weaknesses of an approach is to adopt aspects based on a hybrid solution, which in this work considers: taking advantage of the different types of RS; using problem-solving techniques; and combining resources from different sources to compose a unified metric to be used to rank the recommendation by relevance. Among the RS application areas, news recommendation stands out, being used by a heterogeneous public, ample and demanding by relevance. In this context, the this work shows a hybrid approach to news recommendations built through a architecture implemented to prove the concepts of a recommendation system. This architecture has been validated by using a news corpus and by performing an online experiment. Through the experiment it was possible to observe the architecture capacity related to the requirements of a news recommendation system and architecture also related to privilege recommendations based on similarity, popularity, diversity, novelty and serendipity. It was also observed an evolution in the indicators of reading, likes, acceptance and serendipity as the system accumulated a history of preferences and solutions. Through the analysis of the unified metric for ranking, it was possible to confirm its efficacy when verifying that the best classified news in the ranking was the most accepted by the users