On the estimation of frequent itemsets for data streams: theory and experiments

In this paper, we devise a method for the estimation of the true support of itemsets on data streams, with the objective to maximize one chosen criterion among {precision, recall} while ensuring a degradation as reduced as possible for the other criterion. We discuss the strengths, weaknesses and range of applicability of this method that relies on conventional uniform convergence results, yet guarantees statistical optimality from different standpoints. Categories and Subject Descriptors: H.3.3 [Information Storage and Retrieval]: Information Search and Retrieval; I.5.3 [Pattern Recognition]: Clustering General Terms: Algorithms.

• Type: Conference
• Year: 2005
• Where: CIKM
• Authors: Pierre-Alain Laur, Richard Nock, Jean-Emile Symphor, Pascal Poncelet
• Keywords: CIKM 2005, Conventional Uniform Convergence, Data Streams, Guarantees Statistical Optimality