Skip to content

Commit 8a6b005

Browse files
swilly22swilly22
committed
added cdlp
1 parent e3752a8 commit 8a6b005

File tree

2 files changed

+185
-0
lines changed

2 files changed

+185
-0
lines changed

‎algorithms/cdlp.md

Lines changed: 181 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,181 @@
1+
---
2+
title: "Community Detection using Label Propagation (CDLP)"
3+
description: "Community Detection using Label Propagation (CDLP)"
4+
parent: "Algorithms"
5+
---
6+
7+
# Community Detection using Label Propagation (CDLP)
8+
9+
## Overview
10+
11+
The Community Detection using Label Propagation (CDLP) algorithm identifies communities in networks by propagating labels through the graph structure.
12+
Each node starts with a unique label, and through iterative propagation, nodes adopt the most frequent label among their neighbors, naturally forming communities where densely connected nodes share the same label.
13+
14+
CDLP serves as a powerful algorithm in scenarios such as:
15+
- Social network community detection
16+
- Biological network module identification
17+
- Web page clustering and topic detection
18+
- Market segmentation analysis
19+
- Fraud detection networks
20+
21+
## Algorithm Details
22+
23+
CDLP initializes by assigning each node a unique label (typically its node ID).
24+
The algorithm then iteratively updates each node's label to the most frequent label among its neighbors.
25+
During each iteration, nodes are processed in random order to avoid deterministic bias.
26+
The algorithm continues until labels stabilize (no changes occur) or a maximum number of iterations is reached.
27+
The final labels represent community assignments, where nodes sharing the same label belong to the same community.
28+
29+
The algorithm's strength lies in its ability to discover communities without requiring prior knowledge of the number of communities or their sizes.
30+
It runs in near-linear time and mimics epidemic contagion by spreading labels through the network.
31+
32+
### Performance
33+
34+
CDLP operates with a time complexity of **O(m + n)** per iteration, where:
35+
- **n** represents the total number of nodes
36+
- **m** represents the total number of edges
37+
38+
The algorithm typically converges within a few iterations, making it highly efficient for large-scale networks.
39+
40+
## Syntax
41+
42+
```cypher
43+
CALL algo.labelPropagation([config])
44+
```
45+
46+
### Parameters
47+
48+
The procedure accepts an optional configuration `Map` with the following parameters:
49+
50+
| Name | Type | Default | Description |
51+
|---------------------|---------|------------------------|----------------------------------------------------------------------------------|
52+
| `nodeLabels` | Array | All labels | Array of node labels to filter which nodes are included in the computation |
53+
| `relationshipTypes` | Array | All relationship types | Array of relationship types to define which edges are traversed |
54+
| `maxIterations` | Integer | 10 | Maximum number of iterations to run the algorithm |
55+
56+
### Return Values
57+
The procedure returns a stream of records with the following fields:
58+
59+
| Name | Type | Description |
60+
|---------------|---------|---------------------------------------------------------------------|
61+
| `node` | Node | The node entity included in the community |
62+
| `communityId` | Integer | Identifier of the community the node belongs to |
63+
64+
## Examples
65+
66+
Let's take this Social Network as an example:
67+
68+
```
69+
(Alice)---(Bob)---(Charlie) (Kate)
70+
| | |
71+
(Diana) | (Eve)---(Frank)
72+
| | | |
73+
(Grace)--(Henry) (Iris)--(Jack)
74+
```
75+
76+
There are 3 different communities that should emerge from this network:
77+
- Alice, Bob, Charlie, Diana, Grace, Henry
78+
- Eve, Frank, Iris, Jack
79+
- Any isolated nodes
80+
81+
### Create the Graph
82+
83+
```cypher
84+
CREATE
85+
(alice:Person {name: 'Alice'}),
86+
(bob:Person {name: 'Bob'}),
87+
(charlie:Person {name: 'Charlie'}),
88+
(diana:Person {name: 'Diana'}),
89+
(eve:Person {name: 'Eve'}),
90+
(frank:Person {name: 'Frank'}),
91+
(grace:Person {name: 'Grace'}),
92+
(henry:Person {name: 'Henry'}),
93+
(iris:Person {name: 'Iris'}),
94+
(jack:Person {name: 'Jack'}),
95+
(kate:Person {name: 'Kate'}),
96+
97+
(alice)-[:KNOWS]->(bob),
98+
(bob)-[:KNOWS]->(charlie),
99+
(alice)-[:KNOWS]->(diana),
100+
(bob)-[:KNOWS]->(henry),
101+
(diana)-[:KNOWS]->(grace),
102+
(grace)-[:KNOWS]->(henry),
103+
(charlie)-[:KNOWS]->(eve),
104+
(eve)-[:KNOWS]->(frank),
105+
(eve)-[:KNOWS]->(iris),
106+
(frank)-[:KNOWS]->(jack),
107+
(iris)-[:KNOWS]->(jack)
108+
```
109+
110+
### Example: Detect all communities in the network
111+
112+
```cypher
113+
CALL algo.labelPropagation() YIELD node, communityId RETURN node.name AS name, communityId ORDER BY communityId, name
114+
```
115+
116+
#### Expected Results
117+
| name | communityId |
118+
|------------|-------------|
119+
| `Alice` | 0 |
120+
| `Bob` | 0 |
121+
| `Charlie` | 0 |
122+
| `Diana` | 0 |
123+
| `Grace` | 0 |
124+
| `Henry` | 0 |
125+
| `Eve` | 2 |
126+
| `Frank` | 2 |
127+
| `Iris` | 2 |
128+
| `Jack` | 2 |
129+
| `Kate` | 10 |
130+
131+
### Example: Detect communities with limited iterations
132+
133+
```cypher
134+
CALL algo.labelPropagation({maxIterations: 5}) YIELD node, communityId
135+
```
136+
137+
### Example: Focus on specific node types
138+
139+
```cypher
140+
CALL algo.labelPropagation({nodeLabels: ['Person']}) YIELD node, communityId
141+
```
142+
143+
### Example: Use only certain relationship types
144+
145+
```cypher
146+
CALL algo.labelPropagation({relationshipTypes: ['KNOWS', 'FRIENDS_WITH']}) YIELD node, communityId
147+
```
148+
149+
### Example: Combine node and relationship filtering
150+
151+
```cypher
152+
CALL algo.labelPropagation({
153+
nodeLabels: ['Person'],
154+
relationshipTypes: ['KNOWS']
155+
}) YIELD node, communityId
156+
```
157+
158+
### Example: Group communities together
159+
160+
```cypher
161+
CALL algo.labelPropagation() YIELD node, communityId
162+
RETURN collect(node.name) AS community_members, communityId, count(*) AS community_size
163+
ORDER BY community_size DESC
164+
```
165+
166+
#### Expected Results
167+
| community_members | communityId | community_size |
168+
|----------------------------------------------------------|-------------|----------------|
169+
| `["Alice", "Bob", "Charlie", "Diana", "Grace", "Henry"]` | 0 | 6 |
170+
| `["Eve", "Frank", "Iris", "Jack"]` | 2 | 4 |
171+
| `["Kate"]` | 10 | 1 |
172+
173+
### Example: Find the largest communities
174+
175+
```cypher
176+
CALL algo.labelPropagation() YIELD node, communityId
177+
RETURN communityId, collect(node) AS nodes, count(*) AS size
178+
ORDER BY size DESC
179+
LIMIT 1
180+
```
181+

‎algorithms/index.md

Lines changed: 4 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -45,3 +45,7 @@ For path expressions like `shortestPath()` used directly in Cypher queries, refe
4545

4646
- **[WCC (Weakly Connected Components)](./wcc.md)**
4747
Finds weakly connected components in a graph, where each node is reachable from others ignoring edge directions.
48+
49+
- **[CDLP (Community Detection Label Propegation)](./cdlp.md)**
50+
Detects communities in a network, by propagating labels through the graph structure.
51+

0 commit comments

Comments
 (0)