Algorithm Categories

Introduction

Before diving into projections and hands-on work, let’s map the landscape.

GDS provides dozens of algorithms across five categories. Each answers a different type of question.

By the end of this lesson, you’ll be able to:

Identify the five categories of GDS algorithms and their purposes
Match analytical questions to the appropriate algorithm category
Recognize that algorithms compute scores—interpretation requires domain knowledge
Understand basic graph requirements for different algorithm types

Category	Question	Example Algorithms
Centrality	Which nodes are most important?	PageRank, Degree, Betweenness
Community Detection	What groups exist?	Louvain, Leiden, WCC
Similarity	Which nodes are alike?	Node Similarity, KNN
Pathfinding	What’s the best route?	Dijkstra, Yen’s K-Shortest
Node Embeddings	How do I represent this for ML?	FastRP, Node2Vec

Start with your question:

A note on mindset: algorithms don’t "find" fraud or "discover" communities.

They compute scores and assignments based on graph structure. You decide what those results mean.

Whether a cluster represents a fraud ring, a department, or noise depends on your domain knowledge.

Some algorithms have specific graph requirements:

You’ll learn to configure projections for these requirements.

In the next lessons, you’ll learn:

GDS algorithms fall into five categories:

Your analytical question determines which category to use.