Discovering influential data objects over time

Abstract

In applications such as market analysis, it is of great interest to product manufacturers to have their products ranked as highly as possible for a significant number of customers. However, customer preferences change over time, and product manufacturers are interested in monitoring the evolution of the popularity of their products, in order to discover those products that are consistently highly ranked. To take into account the temporal dimension, we define the continuous influential query and present algorithms for efficient processing and retrieval of continuous influential data objects. Furthermore, our algorithms support incremental retrieval of the next continuous influential data object in a natural way. To evaluate the performance of our algorithms, we conduct a detailed experimental study for various setups. © 2013 Springer-Verlag.