增量式个性化推荐系统研究

互联网个性化推荐系统(Internet personal recommender systems)是根据用户的兴趣推荐最相关的互联网信息给用户的系统。在网上信息过载矛盾越来越严重、用户信息检索的个性化需求日益增强的现状下，推荐系统已经在搜索引擎、电子商务、网上社区等互联网关键应用中起到了关键性的作用，并且越来越受到重视。 然而，在大型网站上部署一个成熟推荐系统的代价依然很大，需要大量的计算和存储资源，推荐的准确性也依然有很大提升空间和需求，这就为推荐系统的研究提供了很多挑战。在这些挑战中推荐算法的准确性和可扩展性一直是该领域最为关注的两个问题，所谓推荐的准确性是指推荐的信息中用户真正感兴趣的比例，而可扩展性指的是系统能否在可容忍的时间和空间复杂度内处理海量的数据。如何在提高算法推荐准确性的同时增强算法的可扩展性是推荐系统改进的主要研究目标。然而，目前学术界的研究更多侧重于提高推荐算法的准确性，而对于可扩展性，很多准确性很高的算法由于需要比较复杂的计算，处理大规模动态数据的能力往往比较有限，并且它们的评测实验中并没有将可扩展性纳入到评价范畴，导致这些算法目前还很难在工业界大规模应用。 本论文的研究试图解决这一问题。通过在推荐算法中借鉴增量学习(Incremental learning)的思想，即考虑最新的训练数据来更新原有的机器学习模型，不需要或仅需要参考部分旧的训练数据，相对于使用全部数据也即批量的处理方式，增量式改进可以大大降低模型更新的复杂度，从而可以大幅度提高推荐算法在遇到新的训练数据时推荐模型更新的效率，降低计算代价，使得推荐模型的更新可以更加及时，进而提高推荐结果的准确性。具体来说，我们在提出了两种新的增量式协同过滤算法的同时，采用增量式学习的方法对目前准确性最好的若干推荐算法进行加速，特别是提高这些算法面对新的训练数据的更新模型的速度和效率，从而为这些算法的大规模的应用提供了可能。另一方面，新的训练数据包含了最新的用户兴趣，因此相对于旧的训练数据，算法在做更新时应给予更高的权重，这样才能做到推荐的结果在考虑到用户长期兴趣的同时，特别考虑用户近期的兴趣，从而使得推荐结果更加准确。这两方面归纳起来，我们旨在通过增量式学习使得推荐算法在更新时更加高效和精确，真正适用于互联网上海量数据的推荐，同时对其他增量式推荐系统方面的研究也具有借鉴意义。我们的改进工作主要包括以下几个方面： 基于主题模型的增量式推荐算法。主题模型，特别是概率隐含主题模型(PLSA)是一种广泛应用于推荐系统的主流方法，在文本推荐、图像推荐以及协同过滤推荐领域都有着很好的推荐效果。目前制约PLSA算法取得更大成功的重要因素就是PLSA算法更新的复杂度过高，使得学习模型的更新只能做批量式处理，这样就导致推荐的时效性不高，也没有办法体现用户的最新的兴趣和整体的最新动态。我们提出了一种增量式学习方法，可以应用于文本分析领域和协同过滤领域，当有新的训练数据到来时，对于基于文本的推荐，增量式更新方法仅寻找最相关的用户和文本以及涉及到的单词进行主题分布的更新，并给予新的文本以更高权重；对于协同过滤，我们的方法仅对当前用户所评分过得物品以及当前物品所涉及的用户进行更新，大大降低了更新的运算复杂度，提高了新数据在推荐算法中所占的权重，使得推荐更加准确、及时。我们的算法在天涯问答文本数据集上和MovieLens电影推荐数据集、Last.FM歌曲推荐数据集、豆瓣图书推荐数据集等协同过滤数据集上取得了很好的效果。 基于蚁群算法(Ant colony algorithm)的协同过滤推荐方法。受到群体智能(Swarm intelligence)算法的启发，我们提出了一种类似于蚁群算法的协同过滤推荐方法——Ant Collaborative Filtering，初始化阶段该方法给予每个用户或一组用户以全局唯一的单位数量的信息素，当用户对物品评分或者用户表示对该物品感兴趣时，用户所携带的信息素相应的传播到该物品上，同时该物品上已有的信息素（初始化为0）也会相应的传播给该用户；此外，用户和物品所携带的信息素会随着时间的推移有一定速率的挥发，通过挥发机制，可以在推荐时更重视用户近期的兴趣；推荐阶段，按照用户和物品所携带的信息素的种类和数量，我们可以得到相应的相似度，进而通过经典的相似度比较的方法来进行推荐。基于蚁群的协同过滤方法的优势在于可以有效的降低训练数据中的稀疏性，并且推荐算法可以实时的进行更新和推荐，同时考虑了用户兴趣随着时间的变化。我们在MovieLens电影评分、豆瓣书籍推荐、Last.FM音乐推荐数据集上验证了我们的方法。最后，我们建立了一个互联网新闻推荐系统，该系统以Firefox插件形式实现，自动采集用户浏览兴趣和偏好，后端使用不同的推荐算法推荐用户感兴趣的新闻给用户。 基于联合聚类(Co-clustering)的两阶段协同过滤方法。聚类(Clustering)是一种缩小数据规模、降低数据稀疏性的有效方法。对于庞大而稀疏的协同过滤训练数据来说，聚类是一种很自然事实上也的确很有效的预处理方法。因此我们提出了一种两阶段协同过滤框架：首先通过我们提出的一种联合聚类的方法，将原始评分矩阵分解成很多维度很小的块，每一块里面包含相似的用户对相似的物品的评分，然后通过矩阵拟合的方法（我们使用了非负矩阵分解NMF和主题模型PLSA）来对这些小块中的未知评分进行预测。当用户新增了对于某物品的一条评分，我们仅需要更新该用户或该物品所处的数据块进行重新评分预估，大大加快了评分预估的速度。我们在MovieLens电影评分数据集上验证了该算法的效果。 本文的研究成果不仅可以直接应用于大型推荐系统中，而且对于增量式推荐系统的后续研究也具有一定的指导意义。首先基于PLSA的增量式推荐算法对于其他基于图模型的推荐系统具有借鉴价值，其次蚁群推荐算法为一类新的、基于群体智能(Swarm intellignece)的协同过滤算法做出了有价值的探索，最后我们提出的两阶段协同过滤框架对于提高推荐算法的可扩展性和更新效率提出了一个通用的有效解决方案。 推荐系统是一个无止尽的优化的过程，除了推荐精度的不断提高之外，推荐算法的性能随着互联网上数据量的增加也需要进一步提高，增量式学习无疑是提高推荐算法更新速度最重要的方法，本文的研究为这一方向提供了参考。

Finding optimal alternatives based on efficient comparative preference inference

Choosing the right or the best option is often a demanding and challenging task for the user (e.g., a customer in an online retailer) when there are many available alternatives. In fact, the user rarely knows which offering will provide the highest value. To reduce the complexity of the choice process, automated recommender systems generate personalized recommendations. These recommendations take into account the preferences collected from the user in an explicit (e.g., letting users express their opinion about items) or implicit (e.g., studying some behavioral features) way. Such systems are widespread; research indicates that they increase the customers' satisfaction and lead to higher sales. Preference handling is one of the core issues in the design of every recommender system. This kind of system often aims at guiding users in a personalized way to interesting or useful options in a large space of possible options. Therefore, it is important for them to catch and model the user's preferences as accurately as possible. In this thesis, we develop a comparative preference-based user model to represent the user's preferences in conversational recommender systems. This type of user model allows the recommender system to capture several preference nuances from the user's feedback. We show that, when applied to conversational recommender systems, the comparative preference-based model is able to guide the user towards the best option while the system is interacting with her. We empirically test and validate the suitability and the practical computational aspects of the comparative preference-based user model and the related preference relations by comparing them to a sum of weights-based user model and the related preference relations. Product configuration, scheduling a meeting and the construction of autonomous agents are among several artificial intelligence tasks that involve a process of constrained optimization, that is, optimization of behavior or options subject to given constraints with regards to a set of preferences. When solving a constrained optimization problem, pruning techniques, such as the branch and bound technique, point at directing the search towards the best assignments, thus allowing the bounding functions to prune more branches in the search tree. Several constrained optimization problems may exhibit dominance relations. These dominance relations can be particularly useful in constrained optimization problems as they can instigate new ways (rules) of pruning non optimal solutions. Such pruning methods can achieve dramatic reductions in the search space while looking for optimal solutions. A number of constrained optimization problems can model the user's preferences using the comparative preferences. In this thesis, we develop a set of pruning rules used in the branch and bound technique to efficiently solve this kind of optimization problem. More specifically, we show how to generate newly defined pruning rules from a dominance algorithm that refers to a set of comparative preferences. These rules include pruning approaches (and combinations of them) which can drastically prune the search space. They mainly reduce the number of (expensive) pairwise comparisons performed during the search while guiding constrained optimization algorithms to find optimal solutions. Our experimental results show that the pruning rules that we have developed and their different combinations have varying impact on the performance of the branch and bound technique.

La Web Semántica y los archivos de televisión: estado de la cuestión

Se analizan y describen las principales líneas de trabajo de la Web Semántica en el ámbito de los archivos de televisión. Para ello, se analiza y contextualiza la web semántica desde una perspectiva general para posteriormente analizar las principales iniciativas que trabajan con lo audiovisual: Proyecto MuNCH, Proyecto S5T, Semantic Television y VideoActive.

Consensus in organizations:Hunting for the social choice conundrum in APA elections

According to the axiomatic literature on consensus methods, the best collective choice by one method of preference aggregation can easily be the worst by another. Are award committees, electorates, managers, online retailers, and web-based recommender systems stuck with an impossibility of rational preference aggregation? We investigate this social choice conundrum for seven social choice methods: Condorcet, Borda, Plurality, Antiplurality, the Single Transferable Vote, Coombs, and Plurality Runoff. We rely on Monte Carlo simulations for theoretical results and on twelve ballot datasets from American Psychological Association (APA) presidential elections for empirical results. Each of these elections provides partial rankings of five candidates from about 13,000 to about 20,000 voters. APA preferences are neither domain-restricted nor generated by an Impartial Culture. We find virtually no trace of a Condorcet paradox. In direct contrast with the classical social choice conundrum, competing consensus methods agree remarkably well, especially on the overall best and worst options. The agreement is also robust under perturbations of the preference prole via resampling, even in relatively small pseudosamples. We also explore prescriptive implications of our findings.

A hybrid recommendation approach for a tourism system

Many current e-commerce systems provide personalization when their content is shown to users. In this sense, recommender systems make personalized suggestions and provide information of items available in the system. Nowadays, there is a vast amount of methods, including data mining techniques that can be employed for personalization in recommender systems. However, these methods are still quite vulnerable to some limitations and shortcomings related to recommender environment. In order to deal with some of them, in this work we implement a recommendation methodology in a recommender system for tourism, where classification based on association is applied. Classification based on association methods, also named associative classification methods, consist of an alternative data mining technique, which combines concepts from classification and association in order to allow association rules to be employed in a prediction context. The proposed methodology was evaluated in some case studies, where we could verify that it is able to shorten limitations presented in recommender systems and to enhance recommendation quality.

User Profiling and Knowledge Extraction For Tourism

O sector do turismo é uma área francamente em crescimento em Portugal e que tem desenvolvido a sua divulgação e estratégia de marketing. Contudo, apenas se prende com indicadores de desempenho e de oferta instalada (número de quartos, hotéis, voos, estadias), deixando os indicadores estatísticos em segundo plano. De acordo com o “ Travel & tourism Competitiveness Report 2013”, do World Economic Forum, classifica Portugal em 72º lugar no que respeita à qualidade e cobertura da informação estatística, disponível para o sector do Turismo. Refira-se que Espanha ocupa o 3º lugar. Uma estratégia de mercado, sem base analítica, que sustente um quadro de orientações específico e objetivo, com relevante conhecimento dos mercados alvo, dificilmente é compreensível ou até mesmo materializável. A implementação de uma estrutura de Business Intelligence que permita a realização de um levantamento e tratamento de dados que possibilite relacionar e sustentar os resultados obtidos no sector do turismo revela-se fundamental e crucial, para que sejam criadas estratégias de mercado. Essas estratégias são realizadas a partir da informação dos turistas que nos visitam, e dos potenciais turistas, para que possam ser cativados no futuro. A análise das características e dos padrões comportamentais dos turistas permite definir perfis distintos e assim detetar as tendências de mercado, de forma a promover a oferta dos produtos e serviços mais adequados. O conhecimento obtido permite, por um lado criar e disponibilizar os produtos mais atrativos para oferecer aos turistas e por outro informá-los, de uma forma direcionada, da existência desses produtos. Assim, a associação de uma recomendação personalizada que, com base no conhecimento de perfis do turista proceda ao aconselhamento dos melhores produtos, revela-se como uma ferramenta essencial na captação e expansão de mercado.

Rapps Sistema de recomendação de apps de produtividade para Android

Nos últimos anos, a nossa sociedade sofreu alterações significativas ao nível tecnológico que têm vindo a modificar o quotidiano do cidadão e transportaram para a palma da mão um conjunto significativo de tarefas até há poucos anos impensáveis. Atualmente, torna-se possível realizar as mais simples tarefas como, a título de exemplo, efetuar um cálculo matemático, tirar fotografias ou registar numa agenda um compromisso, ou tarefas mais complexas, como por exemplo, escrever ou editar um documento, trabalhar numa folha de cálculo ou enviar um e-mail com um anexo, isto tudo com o recurso a um simples dispositivo móvel, conhecido como smartphone ou tablet. Apesar de existirem diversos tipos de apps que seriam um bom auxílio para o aumento da produtividade dos utilizadores de dispositivos móveis Android, nem todos têm conhecimento das mesmas, pelo que é importante que os utilizadores tenham conhecimentos das vantagens da utilização destes recursos e de tudo o que podem realizar com os seus dispositivos com o objetivo de aumentar a sua produtividade profissional ou pessoal. O presente estudo pretende contribuir para uma análise sobre a potencial utilização das novas tecnologias, mais propriamente estudando e recomendando apps de produtividade. Com este intuito foi criada uma app de recomendação de aplicações de produtividade com recurso a um método de sistemas de recomendação. São apresentados os resultados e as conclusões, com recurso a opiniões de potenciais utilizadores.

LORESA : un système de recommandation d'objets d'apprentissage basé sur les annotations sémantiques

Mémoire numérisé par la Division de la gestion de documents et des archives de l'Université de Montréal

Mures : Un système de recommandation de musique

Pendant la dernière décennie nous avons vu une transformation incroyable du monde de la musique qui est passé des cassettes et disques compacts à la musique numérique en ligne. Avec l'explosion de la musique numérique, nous avons besoin de systèmes de recommandation de musique pour choisir les chansons susceptibles d’être appréciés à partir de ces énormes bases de données en ligne ou personnelles. Actuellement, la plupart des systèmes de recommandation de musique utilisent l’algorithme de filtrage collaboratif ou celui du filtrage à base de contenu. Dans ce mémoire, nous proposons un algorithme hybride et original qui combine le filtrage collaboratif avec le filtrage basé sur étiquetage, amélioré par la technique de filtrage basée sur le contexte d’utilisation afin de produire de meilleures recommandations. Notre approche suppose que les préférences de l'utilisateur changent selon le contexte d'utilisation. Par exemple, un utilisateur écoute un genre de musique en conduisant vers son travail, un autre type en voyageant avec la famille en vacances, un autre pendant une soirée romantique ou aux fêtes. De plus, si la sélection a été générée pour plus d'un utilisateur (voyage en famille, fête) le système proposera des chansons en fonction des préférences de tous ces utilisateurs. L'objectif principal de notre système est de recommander à l'utilisateur de la musique à partir de sa collection personnelle ou à partir de la collection du système, les nouveautés et les prochains concerts. Un autre objectif de notre système sera de collecter des données provenant de sources extérieures, en s'appuyant sur des techniques de crawling et sur les flux RSS pour offrir des informations reliées à la musique tels que: les nouveautés, les prochains concerts, les paroles et les artistes similaires. Nous essayerons d’unifier des ensembles de données disponibles gratuitement sur le Web tels que les habitudes d’écoute de Last.fm, la base de données de la musique de MusicBrainz et les étiquettes des MusicStrands afin d'obtenir des identificateurs uniques pour les chansons, les albums et les artistes.

Who influence the music tastes of adolescents? A study on interpersonal influence in social networks

Research on music information behavior demonstrates that people rely primarily on others to discover new music. This paper reports on a qualitative study aiming at exploring more in-depth how music information circulates within the social networks of late adolescents and the role the different people involved in the process play. In-depth interviews were conducted with 19 adolescents (15-17 years old). The analysis revealed that music opinion leaders showed eagerness to share music information, tended to seek music information on an ongoing basis, and were perceived as being more knowledgeable than others in music. It was found that the ties that connected participants to opinion leaders were predominantly strong ties, which suggests that trustworthiness is an important component of credibility. These findings could potentially help identify new avenues for the improvement of music recommender systems.

Modelling the Human Values Scale from Consumers Transactional Data Bases

The main objective of this paper aims at developing a methodology that takes into account the human factor extracted from the data base used by the recommender systems, and which allow to resolve the specific problems of prediction and recommendation. In this work, we propose to extract the user's human values scale from the data base of the users, to improve their suitability in open environments, such as the recommender systems. For this purpose, the methodology is applied with the data of the user after interacting with the system. The methodology is exemplified with a case study

Adaptation and Personalisation

(with notes in powerpoint)

Methodology to obtain the user's Human Values Scale from Smart User Models

En años recientes,la Inteligencia Artificial ha contribuido a resolver problemas encontrados en el desempeño de las tareas de unidades informáticas, tanto si las computadoras están distribuidas para interactuar entre ellas o en cualquier entorno (Inteligencia Artificial Distribuida). Las Tecnologías de la Información permiten la creación de soluciones novedosas para problemas específicos mediante la aplicación de los hallazgos en diversas áreas de investigación. Nuestro trabajo está dirigido a la creación de modelos de usuario mediante un enfoque multidisciplinario en los cuales se emplean los principios de la psicología, inteligencia artificial distribuida, y el aprendizaje automático para crear modelos de usuario en entornos abiertos; uno de estos es la Inteligencia Ambiental basada en Modelos de Usuario con funciones de aprendizaje incremental y distribuido (conocidos como Smart User Model). Basándonos en estos modelos de usuario, dirigimos esta investigación a la adquisición de características del usuario importantes y que determinan la escala de valores dominantes de este en aquellos temas en los cuales está más interesado, desarrollando una metodología para obtener la Escala de Valores Humanos del usuario con respecto a sus características objetivas, subjetivas y emocionales (particularmente en Sistemas de Recomendación).Una de las áreas que ha sido poco investigada es la inclusión de la escala de valores humanos en los sistemas de información. Un Sistema de Recomendación, Modelo de usuario o Sistemas de Información, solo toman en cuenta las preferencias y emociones del usuario [Velásquez, 1996, 1997; Goldspink, 2000; Conte and Paolucci, 2001; Urban and Schmidt, 2001; Dal Forno and Merlone, 2001, 2002; Berkovsky et al., 2007c]. Por lo tanto, el principal enfoque de nuestra investigación está basado en la creación de una metodología que permita la generación de una escala de valores humanos para el usuario desde el modelo de usuario. Presentamos resultados obtenidos de un estudio de casos utilizando las características objetivas, subjetivas y emocionales en las áreas de servicios bancarios y de restaurantes donde la metodología propuesta en esta investigación fue puesta a prueba.En esta tesis, las principales contribuciones son: El desarrollo de una metodología que, dado un modelo de usuario con atributos objetivos, subjetivos y emocionales, se obtenga la Escala de Valores Humanos del usuario. La metodología propuesta está basada en el uso de aplicaciones ya existentes, donde todas las conexiones entre usuarios, agentes y dominios que se caracterizan por estas particularidades y atributos; por lo tanto, no se requiere de un esfuerzo extra por parte del usuario.