Mining Reliable Models of Associations in Dynamic Databases

Adriano A. VelosoWagner Meira Jr.Márcio L. B. de Carvalho

Most current work on data mining has been focused on devising efficient techniques to In this paper we consider the problem of generating reliable models within the context of association mining, a key data mining task. A model is reliable when it does not lose accuracy and predictive power over time. This is a challenging problem because a model, no matter how accurate, can only predict based on associations mined in the old data. That is,a model requires a semi-stable environment, otherwise it will lose accuracy and predictive power over time. We propose a new approach which, by detecting how the associations are evolving over time, is able to generate more reliable models in dynamic databases. Extensive experimental evaluation on real databases (from e-commerce domain) demonstrates the effectiveness and advantages of the proposed approach.

Caso o link acima esteja inválido, faça uma busca pelo texto completo na Web: Buscar na Web

Biblioteca Digital Brasileira de Computação - Contato:
     Mantida por: