Comparing graph data science libraries for querying and analysing datasets: towards data science queries on graphs

TL;DR

This paper experimentally compares graph data science libraries and management systems for querying and analyzing graphs, highlighting the benefits of combining approaches and parallel execution for data science pipelines.

Contribution

It provides an empirical comparison of analysis tools and management systems, exploring execution strategies and the integration of clustering and prediction models on graphs.

Findings

Different execution strategies impact query performance

Combining analysis tools and management systems is beneficial

Parallel execution platforms enhance data science workflows

Abstract

This paper presents an experimental study to compare analysis tools with management systems for querying and analysing graphs. Our experiment compares classic graph navigational operations queries where analytics tools and management systems adopt different execution strategies. Then, our experiment addresses data science pipelines with clustering and prediction models applied to graphs. In this kind of experiment, we underline the interest in combining both approaches and the interest of relying on a parallel execution platform for executing queries.