Data analytics algorithms in property graph databases: A survey

As data continues to grow in scale, data analytics algorithms play a crucial role in extracting and refining valuable knowledge from data. At the same time, graph databases are essential for efficiently managing network data; among them, the most popular systems are converging towards property graph databases, thereby adding rich semantics to graph modeling. In this survey, we explore how data analytics algorithms are supported by property graph databases. First, we provide a comprehensive description of the main graph data analytics algorithms, by classifying and then explaining forty-five algorithms. Then, we examine the twenty most popular graph databases, based on an externally provided usage ranking, and map the data analytics algorithms to them, discovering that only ten property graph databases support some data analytics algorithms. The outcome of this work provides an indication of the coverage of graph data analytics by each property graph database. Finally, to pragmatically guide potential users in choosing the best available solutions for graph data analytics, we evaluate the performance of three property graph databases (Neo4j, Memgraph, and TigerGraph) selected on the basis of usage ranking, data analytics coverage and availability of an open-source version. Our performance evaluation applies to synthetic datasets of different topologies and increasing sizes, and to five real-world graphs that exhibit different network features.