CODEWITHSUNDEEP
graphneo4jgraph-databases
abergmeier
abergmeier

Reputation: 14071

Graph Databases with Frontend and advanced features

I am trying to visualize some data. For that there was a smallish directed graph, which I was inserting into Neo4j (their Sandbox made the BTE low enough). The graph reached now 700 nodes with 500 edges (not that big IMO), which now seems to be too much for Neo4js frontend. Any time I try to display more than 200 edges, the HTML UI gets unresponsive with no feedback.

Now I am wondering whether there are:

  1. UIs that connect to Neo4j which can visualize graphs with more than 1000 nodes/1000 edges.
  2. A graph database (plus UI), which is proven to handle bigger datasets
  3. Any UI, which supports editing

I am not caring so much about performance but rather about visualizing my data.

Upvotes: 2

Views: 1800

Answers (1)

FrobberOfBits
FrobberOfBits

Reputation: 18022

There are a number of different products and approaches that can visualize big graphs from neo4j, check the link. Another one not mentioned behind that link is gephi, which is a desktop based tool but still works.

I would warn you against general force-directed layouts of a few thousand nodes at a time, because they're rarely what you want. Generally when you're trying to visualize big graph data sets, you're going to want to think about how to lay things out differently. For example, if your dataset is temporally ordered you might be better off with a timeline than a graph. Also keep in mind that neo4j can return tabular results from cypher queries, so with a tabular result you can put that data into any existing visualization tool.

Graph visualization is a really deep topic and it's hard to give a single solid answer without more on your requirements. I will say though that in data viz, usually you really just don't want to see thousands of data points, pretty much ever. When you do want to see that much data, it's usually because it's clustered together into clouds so that what you're really seeing is the clusters, not the points. So you might consider lumping many of your data points into clusters, and then trying to visualize THOSE rather than the raw nodes/edges.

Upvotes: 5

Related Questions