Module recap

What you’ve learned

You now understand how to work with GDS algorithms at a fundamental level.

Algorithm categories

GDS provides 40+ algorithms across five categories:

Centrality - Identify important or influential nodes (Degree, PageRank, Betweenness)
Community Detection - Find natural clusters and groups (Louvain, Leiden, Label Propagation)
Pathfinding - Discover optimal routes through networks (Dijkstra)
Similarity - Find structurally similar nodes (Node Similarity, FastRP)
Embeddings - Create vector representations of nodes (FastRP, GraphSAGE)

Five execution modes

You learned that every algorithm supports five modes:

Stats - Get summary statistics without individual results
Stream - View results directly without storing them
Mutate - Store results in the projection for chaining algorithms
Write - Persist results back to the database
Estimate - Check memory requirements before running

Each mode serves a different purpose in your workflow. You practiced using stats and stream for exploration, and learned when to use mutate for algorithm chaining.

Working with Algorithm Results

You learned to use GDS utility functions to work with algorithm outputs:

gds.util.asNode() - Convert GDS node IDs to database nodes for property access
gds.util.nodeProperty() - Access specific properties directly
gds.util.asNodes() - Convert multiple node IDs at once

These functions are essential when working in stream mode, allowing you to retrieve node names, properties, and other attributes from algorithm results.

Reading GDS documentation

You can now independently learn any GDS algorithm by understanding how to read the documentation:

Introduction - Understand the algorithm’s purpose and when to use it
Syntax - Find required parameters and modes
Configuration - Explore customization options
Attributes - Check whether the algorithm supports directed, undirected, weighted, or heterogeneous graphs

Algorithm configuration

You practiced customizing algorithms with both universal and algorithm-specific configurations:

Universal options: nodeLabels, relationshipTypes, concurrency
Orientation options: undirectedRelationshipTypes for making relationships bidirectional
Weight options: relationshipWeightProperty to use weighted relationships
Algorithm-specific options: dampingFactor for PageRank, gamma and theta for Leiden

You ran algorithms with different configurations to see how they affect results:

Directed vs. Undirected: Compared PageRank on directed and undirected actor networks
Unweighted vs. Weighted: Ran Leiden with and without rating weights
Configuration tuning: Experimented with Leiden’s resolution parameters

Relationship aggregation

You learned how to aggregate multiple relationships during projection using count(r):

What it does: Collapses parallel relationships into single weighted relationships
Why it matters: Reduces graph size while preserving connection strength
When to use it: When multiple relationships represent repeated interactions (actor collaborations, shipment frequency)

You practiced building aggregated projections for actor collaboration networks, director-actor networks, and user-genre networks.

Projection modeling

You learned to design projections based on analytical questions:

Start with the question: "Which actors are most influential?" or "Which directors work in the same circles?"
Choose node types: Monopartite (Actor → Actor) or Bipartite (User → Movie)
Determine relationships: Direct, inferred, weighted, or aggregated
Select the algorithm: Match algorithm assumptions to your projection structure

You practiced designing projections for centrality, community detection, and influence questions, understanding that the same data can be projected different ways depending on what you’re trying to measure.

The skills you have

You can now:

Understand what GDS algorithms do and when to use them
Read GDS documentation independently to learn new algorithms
Choose the right execution mode for your workflow
Use utility functions to access node properties from results
Configure algorithms for directed, undirected, and weighted relationships
Aggregate relationships during projection to create meaningful weights
Design projections that match your analytical questions
Run complete GDS workflows from question to projection to analysis

What’s Next: Real-World Applications

Congratulations! You’ve completed Getting Started with Graph Data Science.

You now have a strong foundation in GDS fundamentals. You understand how to:

Create projections tailored to analytical questions
Choose appropriate algorithms for different problem types
Configure algorithms with orientation, weights, and parameters
Interpret and work with algorithm results
Read GDS documentation independently

Ready to try these skills in a real-world scenario?

Continue with Applied Algorithms in GDS, where you’ll explore real-world use cases:

Manufacturing Optimization - Root cause analysis using centrality and community detection
Fraud Detection - Network-based fraud identification
Supply Chain Logistics - Route optimization with pathfinding algorithms
Citation Networks - Find influential papers within a large graph

Each module demonstrates when and why specific algorithms work, showing you not just how to run them, but when and why professionals choose specific approaches for complex problems.

Summary

You’ve completed the foundational course in Graph Data Science! You’ve learned algorithm categories, execution modes, utility functions, configuration, relationship aggregation, and projection modeling.

Continue your journey with Applied Algorithms in GDS to see these techniques solve real-world problems across industries.

Get started with the Graph Data Science library