Title:	EXTRACTION OF EVOLUTIVE FEATURES IN RECOMMENDATION DOMAINS
Author(s):	Fernando Mourão, Alan Cardoso, Leonardo Rocha, Wagner Meira Jr.
ISBN:	978-989-8533-01-2
Editors:	Bebo White, Pedro Isaías and Flávia Maria Santoro
Year:	2011
Edition:	Single
Keywords:	Recommendation, Characterization, Temporal Analysis
Type:	Full Paper
First Page:	404
Last Page:	412
Language:	English
Cover:
Full Contents:	click to dowload
Paper Abstract:	Recommender Systems (RSs) have become increasingly important tools for various commercial applications on the Web. A primordial and very recurrently studied challenge in these systems consists in the temporal evolution of users and do-mains. However, little understanding exists about how this evolution actually emerges. For such, a characterization me-thodology for evolutive behavior was recently proposed in the literature [Cardoso, A., et al. 2011], composed by three steps, which has proved to be effective and useful in a real recommendation scenario. In this work we extend this metho-dology by adding a fourth step, that we named Users Taste Analysis, which aims to analyze what the users consume in the system across the time, providing an even more robust and deep analysis of evolution. Further, we propose what we call extraction functions to transform the methodology's results in simple and meaningful metrics of evolution (i.e., evo-lutive features), defining a straight way to exploit all acquired knowledge into traditional RSs. In order to validate our extract functions, we choose the music recommendation scenarios, more specifically the Last.Fm application. Our results, at first, show that our methodology extension also provides a very rich source of information for describing recommenda-tion domains over time. Second, we observe that the extraction functions are able to support an easy interpretation for the methodology results, highlighting some relevant issues not direct identified by them.
	Go Back