Diversified set monitoring over distributed data streams

Abstract

Data monitoring over distributed streams is a fundamental problem, as represented by modern applications, e.g., sensor network and financial data monitoring. Such applications need a technique which continuously monitors user-requiring data and achieves not only time and space efficiencies but also communication efficiency. In addition, result diversification is also required to increase user satisfaction, thus has been receiving significant attention recently. This motivates us to consider a problem of monitoring k-diverse data over distributed streams. Result diversification is well known to be NP-hard, so the natures of NP-hardness and dynamic distributed data bring non-trivial challenges, e.g., impracticably of centralized approaches. In this paper, we propose a novel algorithm that monitors k-diverse data with time, space, and communication efficiencies. The results of our experiments using both real and synthetic data confirm the effectiveness of our algorithm.