##Set up Neo4j Enterprise (Docker / VM).
docker run \
-p 7474:7474 -p 7687:7687 \
-e NEO4J_AUTH=neo4j/s3cr3t\
-v $PWD/data:/data -v $PWD/plugins:/plugins \
--name neo4j-apoc \
-e NEO4J_apoc_export_file_enabled=true \
-e NEO4J_apoc_import_file_enabled=true \
-e NEO4J_apoc_import_file_use__neo4j__config=true \
-e NEO4JLABS_PLUGINS=\[\"apoc\"\] \
--env NEO4JLABS_PLUGINS='["graph-data-science"]'\
neo4j
Link to the graph:
https://arrows.app/#/local/id=CGvPhHgNFdKpruz0pgS1
There are four activities can be identified given the supplied dataset, that is Person node can have any of the following four activity nodes:
- Work
- Education
- Trip
- Transaction
In this exercise, I created four nodes for different activities, an alternative way, given the small amount of data, could be creating an activity node, with a type property set as work, education etc.
To reduce the amount of duplicated data, I further introduced the country and year nodes, as these data are presented in all tables. Based on the assumed graph, I intend to answer the following use cases:
Fact questions:
- Has an individual worked/educated/traveled/shopped ?
- If any activity happened, what is the location?
- Given a particular year, what a person has done?
- How many times a person has shopped?
Inference question:
- Does more educated person shopped more?
After code execution 103 nodes will be created:
- 30 Transaction nodes
- 3 Person nodes
- 15 Trip nodes
- 18 Country nodes
- 12 Education nodes
- 6 Work nodes
- 19 Year nodes
Please refer to neo4j_test.py for the import application.
Case one: Has Anil Kumar received any education?
MATCH (n:Person)-[:HAS_ACTIVITY {type:'education'}]-(e) WHERE n.name = 'Anil Kumar' RETURN e
Answer: Anil has received five qualifications
Case two: When and where did Anil Kumar receive her university education?
MATCH (n:Person)-[:HAS_ACTIVITY {type:'education'}]-(e:Education)-[:EVENT_LOCATION_AT]-(c:Country)
MATCH (n:Person)-[:HAS_ACTIVITY {type:'education'}]-(e:Education)-[:EVENT_ACTION_YEAR_START]-(y:Year)
WHERE n.name = 'Anil Kumar'
AND e.nameofinstitution CONTAINS 'University'
RETURN c.name, y.year
Answer: 1992 in Vietnam.
Case three: What has Anil Kumar did during 1990-2000 (inclusive)?
MATCH (n:Person)-[r:HAS_ACTIVITY]-(e)-[:EVENT_ACTION_YEAR_START]-(y:Year)
WHERE n.name = 'Anil Kumar'
AND 1990<= y.year <=2000
RETURN labels(e)[0], y.year
Answer: Anil Kumar started education in 1992 and started work in 1996.
Case four: How often does Anil Kumar shop?
MATCH (n:Person)-[r:HAS_ACTIVITY {type:'shopping'}]-(e)-[:EVENT_ACTION_YEAR_START]-(y:Year)
WHERE n.name = 'Anil Kumar'
RETURN y.year,count(r)
Answer: 10 times in year 2021.
When looking only at the location based similarity, Alice Gan and Anil Kumar are mostly similar
CALL gds.graph.project(
'myGraph3',
['Country','Person'],
'PERSON_ACTION_LOCATION'
)
CALL gds.beta.closeness.stream('myGraph3')
YIELD nodeId, score
RETURN gds.util.asNode(nodeId).name AS name, score
ORDER BY score DESC
Person1 Person2 similarity
"Alice Gan" "Anil Kumar" 0.181
"Alice Gan" "Ariff Johan" 0.091
"Anil Kumar" "Ariff Johan" 0.083
The country they visited most (apart from their home country) is Vietnam.
MATCH (p:Person) - [:PERSON_ACTION_LOCATION] - (c:Country)
WHERE p.name = "Alice Gan" or p.name = "Anil Kumar"
Return c.name, count(c.name)
Anil Kumar and Ariff Johan are, on the other hand, similar in terms of the activity year.
Person1 Person2 similarity
"Anil Kumar" "Ariff Johan" 0.666
"Alice Gan" "Ariff Johan" 0.636
"Anil Kumar" "Alice Gan" 0.538
After including both time and location, we can see all three individuals are quite similar. This is because all of them has studied at Smart National University of Vietnam during 1992 and 1995 for the same degree. They also visited Vietnam in September 2021.
CALL gds.graph.project(
'myGraph',
['Country','Person','Year'],
['PERSON_ACTION_LOCATION','PERSON_ACTION_YEAR_START']
)
CALL gds.beta.closeness.stream('myGraph')
YIELD nodeId, score
RETURN gds.util.asNode(nodeId).name AS name, score
ORDER BY score DESC
Person1 Person2 similarity
"Alice Gan" "Anil Kumar" 0.375
"Anil Kumar" "Ariff Johan" 0.375
"Alice Gan" "Ariff Johan" 0.363
The linkage between the transaction data and the travel data gives further evidence that all three individuals visited Vietnam together on 2021-09-26
MATCH (p:Person)-[r:HAS_ACTIVITY {type:'shopping'}]-(e)-[:EVENT_LOCATION_AT]-(c)
WHERE c.name = "Vietnam" and e.date >= date("2021-09-26")
RETURN p.name, e.merchant, sum(e.amount)
name e.merchant sum(e.amount)
"Alice Gan" "GRAND COCONUT HOTEL, VIETNAM" 1230.0
"Ariff Johan" "GRAND COCONUT HOTEL, VIETNAM" 1230.0
"Anil Kumar" "GRAND COCONUT HOTEL, VIETNAM" 1230.0
match (e) - [:EVENT_ACTION_YEAR_START] - (y:Year)
match (p:Person) - [:PERSON_ACTION_YEAR_START] - (y:Year)
match (p:Person) - [:HAS_ACTIVITY] - (e)
match (p:Person) - [:PERSON_ACTION_LOCATION] - (c)
match (e) - [:EVENT_LOCATION_AT] - (c)
where c.name = "Vietnam"
return y,p,e,c
Although they graduated from the same place, they work in the same country as their citizenship.
MATCH (p:Person)
CALL {
WITH p
MATCH (p:Person)-[:PERSON_ACTION_LOCATION {type:'work'}]-(c)
RETURN c.name AS work_country
}
CALL {
WITH p
MATCH (p:Person)-[:PERSON_ACTION_LOCATION {type:'education'}]-(c)
RETURN c.name AS education_country
}
RETURN p.name, work_country
p.name work_country education_country
"Alice Gan" "Malaysia" "Malaysia"
"Alice Gan" "Malaysia" "Vietnam"
"Anil Kumar" "India" "Malaysia"
"Anil Kumar" "India" "Vietnam"
"Ariff Johan" "Singapore" "Vietnam"
"Ariff Johan" "Singapore" "Singapore"
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent