Projected Graph Recipe

The Projected graph recipe creates an edge-list dataset from a bipartite dataset.

For example, if your input dataset contains customer / product pairs, you can create a graph of products linked by customers. The output contains product-to-product pairs, and can optionally include a weight column with the number of shared linking values.

This recipe is a dataset-to-dataset compatibility recipe. It does not require a Visual Graph graph folder.

Input / Output

Input

• Input dataset: Dataset containing the bipartite pairs.

Output

• Output dataset: Dataset containing the projected edge list.

Settings

Create graph of

Select the column containing the entity type to project. Values from this column become the nodes in the output edge list.

Linked by

Select the column containing the linking entity type. Two projected nodes are connected when they share at least one value in this column.

The two selected columns are interpreted by their roles in the recipe. The same literal value can appear in both columns without being treated as the same entity.

Weighted graph

Enable this option to add a weight column. The weight is the number of shared linking values for each projected pair.

The recipe loads the selected columns in memory. Very large bipartite datasets can require substantial RAM.