Neo4j has this great IDE-a: How about we stuff all our graph workspace, database, algorithms and visualisation wizardry in one place?

Graph graph graph egg and graph; graph graph graph graph graph graph baked beans graph graph graph

Graph database slinger Neo4j is putting out a suite of tools aimed at helping data scientists be more productive using graph analytics techniques.

It promises a native graph analytics workspace, graph database, scalable graph algorithms and graph visualisation all in one environment.

Data science and machine learning are treading a path similar to that taken by software development. Whereas way back in history, programmers might mix and match tools and techniques, since the early 1990s, most have used an integrated development environment (IDE) to make the process more productive and consistent.

Data scientists are still pulling tools from all over the place, but some companies, notably H2O.ai and DataRobot, have tried to bring the most useful into a single product. Azure and Google Cloud Platform have similar environments.

But, according to Neo4j, none address graph databases – that is, databases with a structure specifically designed to describe network structures and interactions, such as social media or traffic flow. So, it has made one that does.

Neo4j for Graph Data Science offers a ready-to-hand library of algorithms that the vendor said were optimised to run over tens of billions of nodes and relationships. It includes production features, such as deterministic seeding which provides starter values and consistent results for reproducible machine learning workflows. It also contains its own in-memory graph database for training and test data.

The firm touts a "friendly data science experience with logical memory management, intuitive API and extensive documentation and guides."

To see the results of an analysis, Neo4j is including its graph visualisation tool Bloom, which is also available as a standalone product.

Amy Hodler, director of graph analytics and AI programmes at Neo4j, said Bloom had been tweaked to make it more useful in predictive analytics. "We allow Bloom to visualize based on the algorithm results," she said. "You can always look at your graph, but we did things like add a larger node size for a bigger PageRank score, for example."

She said the tool would complement environments such as H2O.ai and DataRobot.

"Graph Data Science is not intended to replace those tools," Hodler added. "It does not have some of the algorithms and the functions that those platforms have and they don't have any graph functions. But in the future we definitely want to do some integration so that you can do graph feature engineering in Neo4j and then use those features in other platforms."

It's a good step to make a single environment to boost productivity in graph analytics. But ultimately data scientists will want to do analytics on graph and other data structures within the same environment. Whether that will be within Neo4j's tool or one of the other popular autoML tools will be for the market to decide. ®