IRISA Participation in JRS 2012 Data-Mining Challenge: Lazy-Learning with Vectorization

Vincent Claveau (1)

1 TEXMEX - Multimedia content-based indexing

In this article, we report on our participation in the JRS Data-Mining Challenge. The approach used by our system is a lazy- learning one, based on a simple k-nearest-neighbors technique. We more specifically addressed this challenge as an opportunity to test Informa- tion Retrieval (IR) inspired techniques in such a data-mining framework. In particular, we tested different similarity measures, including one called vectorization that we have proposed and tested in IR and Natural Lan- guage Processing frameworks. The resulting system is simple and efficient while offering good performance.

Domaines: Recherche d'information [cs.IR]; Apprentissage [cs.LG]

Vincent Claveau. IRISA Participation in JRS 2012 Data-Mining Challenge: Lazy-Learning with Vectorization. JRS - Data Mining Competition: Topical Classification of Biomedical Research Papers, special event of Joint Rough Sets Symposium, Sep 2012, Chengdu, China. ⟨hal-00760145⟩