StreamPref: a query language for temporal conditional preferences on data streams

Abstract

Over recent years, several studies regarding preference queries over data streams have been developed in database and artificial intelligence research fields. Preference queries are useful in many decision making application areas, such as e-commerce, financial analysis, and content personalization. In this article, we explore new aspects of temporal conditional preference queries (tcp-queries) for the StreamPref query language. Tcp-queries allow the user to express how past instants of a data stream can influence the preference of a user at a present instant. In order to increase the utility of the StreamPref query language, we propose herein new operators that allow dealing with subsequences and filtering of sequences by length. To validate our proposal we present a detailed complexity analysis and an extensive set of experiments with synthetic and real datasets, which corroborate the efficiency of the algorithms and the utility of the new operators.