Submitted by William Justo Figueiro (williamjf) on 2015-08-28T17:26:34Z No. of bitstreams: 1 34e.pdf: 3564229 bytes, checksum: 47f535efb73877017d3fc9b78250f707 (MD5) / Made available in DSpace on 2015-08-28T17:26:34Z (GMT). No. of bitstreams: 1 34e.pdf: 3564229 bytes, checksum: 47f535efb73877017d3fc9b78250f707 (MD5) Previous issue date: 2013 / CAPES - Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / Um dos principais desafios de sistemas ubíquos e sensíveis a contexto é a coleta de informações relevantes sobre entidades, e o uso destas informações para compreender e prever seu comportamento. Isto permite que as aplicações adaptem-se às entidades, evitando assim uma sobrecarga de questionamentos e informações à entidade. Este trabalho apresenta o eProfile, um modelo que permite que aplicações registrem as ações de entidades em trilhas e infiram informações de perfil a partir destas trilhas, utilizando interoperabilidade semântica e assim permitindo que diferentes aplicações compartilhem informações em um perfil unificado. Foi desenvolvido um protótipo para a avaliação do modelo, o qual foi integrado com dois diferentes softwares. Foi verificado que é possível enriquecer a geração de perfis de aplicações através da integração com o modelo. As contribuições deste modelo são o uso de trilhas para extrair perfis, a geração de perfis dinâmicos, o gerenciamento de regras de inferência e modelos de entidades dinâmicos e a interoperabilidade semântica do modelo. / Context-aware and ubiquitous systems have the challenge of implicitly collect relevant information about entities, and use this information to understand and predict their behaviour. This allows the applications to adapt themselves to the entities, thus avoiding to overflow them with inquires and information. The analysis of trails, the context-aware history of actions, can further improve the relevance of information. This dissertation proposes a model that allows applications to register entites’ actions in trails and infer profile information from these trails, using semantic interoperability and thus allowing different applications to share information and infer a unified profile. A prototype was developed for the evaluation of the model, and it was integrated with two different softwares. It was verified that was possible to enrich the profile generation of applications through the integration with the modelo. The contributions of this model are the use of trails for extracting profiles, the generation of dynamic profiles, the capability of managing dynamic inference rules for profile generation and the semantic interoperabilty of the model.

Recommendation systems are information filtering systems that help users effectively and efficiently explore large amount of information and identify items of interest. Accurate predictions of users' interests improve user satisfaction and are beneficial to business or service providers. Researchers have been making tremendous efforts to improve the accuracy of recommendations. Emerging trends of technologies and application scenarios, however, lead to challenges other than accuracy for recommendation systems. Three new challenges include: (1) opinion spam results in untrustworthy content and makes recommendations deceptive; (2) users prefer diversified content; (3) in some applications user behavior data may not be available to infer users' preference. This thesis tackles the above challenges. We identify features of untrustworthy commercial campaigns on a question and answer website, and adopt machine learning-based techniques to implement an adaptive detection system which automatically detects commercial campaigns. We incorporate diversity requirements into a classic theoretical model and develop efficient algorithms with performance guarantees. We propose a novel and robust approach to infer user preference profile from recommendations using copula models. The proposed approach can offer in-depth business intelligence for physical stores that depend on Wi-Fi hotspots for mobile advertisement. / Graduate / 0984 / cchenv@uvic.ca