Graphs + Neo4j

A graph G is a collection of entities (nodes) and the relationships (edges) that connect those entities

G = {N, E}

Graphs + Neo4j

A graph database stores your data in a graph.

Tables and foreign keys are nodes and relationships.

• What does math say?

Graphs can be undirected

• What does Neo4j say?

You have to make directed relationships. Forget about them when you make your queries.

Graphs + Neo4j

Nodes can have properties.

Relationships can have properties.

Node labels are the best. Gives us set advantages.

q <- ("MATCH (n:Movie { title: 'Footloose' }) 
      RETURN properties(n) AS properties")
properties <- cypherToList(graph, q)

purrr::map(properties, ~names(.x$properties))

## [[1]]
##  [1] "studio"       "releaseDate"  "imdbId"       "description" 
##  [5] "runtime"      "language"     "title"        "version"     
##  [9] "trailer"      "imageUrl"     "genre"        "tagline"     
## [13] "lastModified" "id"           "homepage"

Neo4j + R

Set your environment

# Connector
library(RNeo4j)
# Manipulate data
library(dplyr)
# Work with graphs
library(igraph)
# Visualize
library(ggplot2)
library(ggthemes)
library(visNetwork)
## If you want to use the Grammar of graphics
# library(ggraph)

graph <- startGraph("http://localhost:7474/db/data/", 
                    username="", 
                    password="")

Neo4j + R

q <- ("MATCH (a:Actor)-[:ACTS_IN]->(m:Movie) 
      RETURN m.genre AS genre, COUNT(m) AS movies 
      ORDER BY movies DESC
      LIMIT 20")
most_movies <- cypher(graph, q)

Neo4j + R

ggplot(most_movies, aes(reorder(genre, -movies), movies)) + 
  geom_bar(stat="identity") +
  theme_tufte() + 
  theme(axis.title.x = element_blank(),
        axis.title.y = element_blank(),
        axis.text.x = element_text(angle = 45, hjust = 1)) +
  ggtitle("Most played genres")

Neo4j + R

q <- ("MATCH (a:Actor)-[:ACTS_IN]->(m:Movie) 
      WHERE m.genre IN ['Comedy', 'Drama', 'Action'] 
      RETURN a.name AS name, m.genre AS genre, COUNT(m) AS count 
      ORDER BY count DESC 
      LIMIT 30")
most_genres <- cypher(graph, q)

Neo4j + R

Visualize

ggplot(most_genres, aes(name, genre)) +
  geom_tile(aes(fill = genre, alpha = count)) + 
  theme_tufte() + 
  theme(axis.title.x = element_blank(),
        axis.title.y = element_blank(),
        axis.text.x = element_text(angle = 45, hjust = 1)) + 
  ggtitle("Actors with most movies by genre ")

RStats

Betweenness centrality

The betweenness centrality for each vertex is the number of these shortest paths that pass through the vertex. It represents the degree of which nodes stand between each other.

\[betweenness(v) = \sum_{x \neq y \in V} \frac{\sigma_{xy}(v)}{\sigma_{xy}}\]

q <- ("MATCH (a1:Actor)-[:ACTS_IN]->(m:Movie),
      (a2:Actor)-[:ACTS_IN]->(m)
      RETURN a1.name AS from, a2.name AS to, COUNT(*) AS weight 
      ORDER BY weight DESC
      LIMIT 10000")
betweenness <- cypher(graph, q)
ig <- graph_from_data_frame(betweenness)

RStats

Betweenness centrality

b <- betweenness(ig)

b[1:5]

##     Larry Simms     Arthur Lake           Daisy Penny Singleton 
##       129.01905        55.02381        64.12143        55.12143 
##   Danny Mummert 
##        97.35476

RStats

Betweenness centrality

V(ig)$label <- NA
V(ig)$size = 4
V(ig)$color = "cyan"

plot(ig, edge.arrow.size = 0.1)