Graph analytics

Recipes to handle graphs and compute graph statistics.

This plugin offers several recipes to handle two kinds of graphs:

  • Edges list (graph specified as a list of source_node / target_node couples)
  • Bipartite graph (two kinds of items like product and transaction, transactions linking products together)

It can transform bipartite graphs into edge lists and extract node statistics from edge lists.

This plugin also provides a template webapp to visualize your graph and perform a community detection.

Plugin Information

Version 0.0.2
Author Dataiku (Thomas Cabrol, Pierre Pfennig)
Released 2016-06-28
Last updated 2018-08-01

How To Use

Compute statistics from list of edges

This recipe takes the list of edges as input, and compute several graph statistics. Beware that some of the possible statistics can be quite long to compute depending of the size of the graph.

Create a projected graph from a bipartite graph

This recipe transforms a bipartite graph (for exemple a list of transactions User – Product) into a list of edges for the projected graph (for exemple the graph of Products). Be aware that his recipe can be memory expensive.

Grap visualization webapp

This webapp gives you tha ability to visualize a graph and perform a community detection. You have to provide the input dataset, the columns corresponding to the source node and the target node, and th column corresponding to the intensity of the interaction (it can be constant to 1). Then, the similarity parameter will give you the ability to take only edges with this minimum number of interactions. The color of the resulting nodes correspond to the different communities.

Get the Dataiku Data Sheet

Learn everything you ever wanted to know about Dataiku (but were afraid to ask), including detailed specifications on features and integrations.

Get the Data Sheet