Top-k Context-Aware Queries on Streams

Preference queries have been largely studied for relational systems but few proposals exist for stream data systems. Most of the existing proposals concern the skyline, top-k or top-k dominating queries, coupled with the sliding-window operator. However, user preferences queries on data streams may be more sophisticated than simple skyline or top-k and may involve more expressive operations on streams. This paper improves the existing work on data stream query-answering personalization by proposing a solution to express and handle contextual preferences together with a large variety of queries including one-shot and continuous queries. It adopts a more expressive preference model supporting context-based preferences, allowing to capture a wide range of situations. We propose algorithms to implement the new preference operators on stream data and validate their performance on a real-world dataset of stock market streams.