Graph Algorithms

Graph algorithms unlock insights hidden in the structure of connected data. Neo4j provides the Graph Data Science (GDS) library — a comprehensive collection of algorithms for pathfinding, centrality, community detection, similarity, and more.

---

The Graph Data Science (GDS) Library

GDS is a Neo4j plugin that provides:

• 65+ graph algorithms across multiple categories
• Graph projections — in-memory copies of your data optimised for algorithm execution
• Execution modes — stream, mutate, write, or estimate
• ML pipelines — link prediction and node classification

Installation

Environment	Method
Neo4j Desktop	Install from the Plugins tab
Neo4j AuraDS	Pre-installed
Docker	Use the GDS-enabled Docker image
Server	Download and install the GDS plugin

---

Graph Projections

Before running algorithms, you project a subgraph into memory:

// Native projection
CALL gds.graph.project(
  'social-graph',        // projection name
  'Person',              // node labels
  'FRIENDS_WITH'         // relationship types
)

Projection options

Option	Description
Node labels	Which node types to include
Relationship types	Which relationship types to include
Properties	Node/relationship properties to include
Orientation	NATURAL, REVERSE, or UNDIRECTED

Drop a projection when finished

CALL gds.graph.drop('social-graph')

---

Algorithm Categories

Category	Purpose	Example Algorithms
Pathfinding	Find routes between nodes	Shortest Path, Dijkstra, A*
Centrality	Identify important nodes	PageRank, Betweenness, Degree
Community Detection	Find clusters and groups	Louvain, Label Propagation, WCC
Similarity	Find similar nodes	Node Similarity, K-Nearest Neighbours
Node Embedding	Generate vector representations	FastRP, Node2Vec, GraphSAGE
Link Prediction	Predict future connections	Common Neighbours, Preferential Attachment

---

Pathfinding Algorithms

Shortest Path (Dijkstra)

Find the shortest weighted path between two nodes:

MATCH (source:City {name: "London"}), (target:City {name: "Paris"})
CALL gds.shortestPath.dijkstra.stream('transport-graph', {
  sourceNode: source,
  targetNode: target,
  relationshipWeightProperty: 'distance'
})
YIELD index, sourceNode, targetNode, totalCost, nodeIds, costs, path
RETURN totalCost, [nodeId IN nodeIds | gds.util.asNode(nodeId).name] AS route

All Shortest Paths

Find all shortest paths between two nodes:

MATCH (source:Person {name: "Alice"}), (target:Person {name: "Dave"})
CALL gds.shortestPath.dijkstra.stream('social-graph', {
  sourceNode: source,
  targetNode: target
})
YIELD totalCost, path
RETURN totalCost, path

Built-in Shortest Path (no GDS required)

Neo4j also has a built-in shortest path function: