Neo4j Project

This project is an exaple of the usage of Neo4j graph database.

Dataset

A dataset available in the Stanford Large Network Dataset Collection (SNAP). It focuses on Massive Open Online Courses (MOOCs) and contains information about user interactions within these online learning platforms.

MOOCs are online courses that are open to anyone and typically attract a large number of participants. This dataset specifically captures user actions and interactions within MOOCs, providing valuable insights into user behavior and engagement.

The dataset contains 3 files:

• mooc_actions.tsv contains information about user actions within the MOOCs.
• mooc_actions_features.tsv contains information about the features of the user actions.
• mooc_actions_labels.tsv contains information about the labels of the user actions.

Data Model

The data model is composed of 2 nodes:

• User node, which represents a user.
• Target node, which represents a target (e.g. a video, a quiz, etc.).

The data model is composed of 1 relationship:

• PERFORMS_ACTION, which represents the action performed by a user on a target.

Load data in Neo4j

To load data into Neo4j we used the main.py file. We used the py2neo library to connect to the database and to load the data. In order to load data we used data structures such as dictionaries and sets.In this way the data loaded faster and we avoided duplicates.

Queries

Show a small portion of your graph database (screenshot)

Count all users, count all targets, count all actions

1. Count all users

MATCH (u:User)
RETURN count(u)

and the result is 7047

2. Count all targets

MATCH (t:Target)
RETURN count(t)

and the result is 97

3. Count all actions

MATCH ()-[r]->()
RETURN count(r) AS actionCount

and the result is 411749

Show all actions (actionID) and targets (targetID) of a specific user (choose one)

MATCH (u:User {id: '1'})
MATCH (u)-[r]->(t:Target)
RETURN r.id AS actionID, t.id AS targetID

and the result is

For each user, count his/her actions

MATCH (u:User)-[r]->()
RETURN u.id AS userID, count(r) AS actionCount

and the result is

For each target, count how many users have done this target

MATCH (u:User)-[r]->(t:Target)
RETURN t.id AS targetID, count(DISTINCT u) AS userCount

and the result is

Count the average actions per user

MATCH (u:User)
OPTIONAL MATCH (u)-[r]->()
WITH u, count(r) AS actionCount
RETURN avg(actionCount) AS averageActionsPerUser

and the result is

Show the userID and the targetID, if the action has positive Feature2

MATCH (u:User)-[r]->(t:Target)
WHERE toFloat(r.feature2) > 0
RETURN u.id AS userID, t.id AS targetID

and the result is:

For each targetID, count the actions with label “1”

MATCH (u:User)-[r]->(t:Target)
WHERE r.label = 1
RETURN t.id AS targetID, count(r) AS actionCount

and the result is:

Authors

• Marios Aintini
• Giorgios Zarkadas