Executando Algoritmos GDS

The Graph Data Science (GDS) library in Neo4j helps us run powerful graph algorithms. To use GDS, you generally follow a three-step process:

• Project: Load a subset of your graph into GDS's optimized in-memory store.
• Execute: Run an algorithm on this in-memory graph.
• Write Back: Optionally, write the results back to your Neo4j database.

Before running an algorithm, GDS needs to know which part of your database graph to analyze. This process is called graph projection.

You define which nodes and relationships, based on their labels and types, should be copied into GDS's fast in-memory representation. This makes algorithm execution much quicker.

Let's project a simple graph containing only nodes with the label Person. We give our projected graph a name, like myGraph.

The gds.graph.project procedure creates this in-memory graph.

More commonly, you'll project both nodes and relationships. Here, we project Person nodes and their KNOWS relationships.

We specify the node label and the relationship type. You can also add properties to the projection.

Once your graph is projected, you can run various algorithms. GDS algorithms often have different execution modes:

• Stream: Returns results directly as a stream of rows. Great for immediate analysis.
• Stats: Returns only summary statistics about the algorithm's run.
• Write: Writes the results back to your Neo4j database as new properties or relationships.

Let's run the PageRank algorithm on our socialGraph in stream mode. PageRank helps identify influential nodes.

We use gds.pageRank.stream and pass the name of our projected graph.

GDS algorithms are highly configurable. You can pass a map of parameters to fine-tune their behavior.

Common parameters include maxIterations, dampingFactor (for PageRank), or concurrency. Always check the GDS documentation for specific algorithm parameters.

Here, we run PageRank with a custom dampingFactor and limit the maxIterations. This gives you more control over the algorithm's execution and convergence.

Often, you'll want to store the results of an algorithm back into your Neo4j database. This allows you to query them later or use them in further analysis.

The write mode of GDS algorithms adds new properties to nodes or creates new relationships based on the algorithm's output.

To write results, use the .write procedure, like gds.pageRank.write. You need to specify the property name where the score will be stored on the original nodes.

After running this, each Person node in your database will have a new property, pageRankScore.

Great job! You've learned the core steps for running GDS algorithms:

• Projecting your graph into GDS memory using gds.graph.project.
• Executing algorithms in stream, stats, or write modes.
• Configuring algorithms with custom parameters.
• Writing back results to your Neo4j database for persistence.

Next, you'll explore advanced GDS pipelines and integration with machine learning!