Query Operators for Comparing Uncertain Graphs

Extending graph models to incorporate uncertainty is important for many applications, including citation networks, disease transmission networks, social networks, and observational networks. These networks may have existence probabilities associated with nodes or edges, as well as probabilities associated with attribute values of nodes or edges. Comparison of graphs and subgraphs is challenging without probabilities. When considering uncertainty of different graph elements and attributes, traditional graph operators and semantics are insufficient. In this paper, we present a prototype SQL-like graph query language that focuses on operators for querying and comparing uncertain graphs and subgraphs. Two interesting operators include ego neighborhood similarity and semantic path similarity. Similarity operators are particularly useful for comparison queries, the focus of this paper. After motivating and describing our operators, we present an implementation of a query engine that uses this query language. This implementation combines a layered and service-oriented architecture and is designed to be extensible, so that simple operators can be used as building blocks for more complex ones. We demonstrate the utility of our query language and operators for analyzing uncertain graphs based on two real world networks, a dolphin observation network and a citation network. Finally, we conduct a performance evaluation of some of the more complex operators, illustrating the viability of these operators for analysis of larger graphs.