Neo4j: Building a Movie-Director Graph - UMD Classes -- Tutorials

This notebook demonstrates a complete end-to-end application that builds a graph database of movies and directors from the Netflix dataset, then visualizes the relationships.

Workflow:

Load the Netflix CSV dataset with Pandas
Clean and preprocess the data
Populate a Neo4j graph database using the Neo4jAPI wrapper
Visualize the director-movie graph with NetworkX

Why graphs for this? Graphs are an intuitive way to represent relationships. Each movie is connected to its director, forming a natural web. Using Neo4j, we can query and explore this data efficiently, and by visualizing it we reveal connections that might otherwise be hard to see in rows and columns.

%load_ext autoreload
%autoreload 2
%matplotlib inline

import logging

import pandas as pd

import helpers.hdbg as hdbg
import helpers.hnotebook as hnotebo
import tutorials.tutorial_neo4j.neo4j_utils as ttneouti

hdbg.init_logger(verbosity=logging.INFO)
_LOG = logging.getLogger(__name__)
hnotebo.config_notebook()

Part 1: Data Loading and Inspection¶

Load the Netflix dataset from artifacts/data/netflix.csv. The dataset contains movie and TV show metadata including title, director, cast, country, and release year.

Sample structure:

| title        | release_year | director      | cast          | country |
|--------------|--------------|---------------|---------------|---------|
| Movie Title  | 2021         | John Doe      | Actor A, ...  | USA     |

# Load the dataset into a Pandas DataFrame.
csv_file = "artifacts/data/netflix.csv"
data = pd.read_csv(csv_file)
_LOG.info("Dataset shape: %s", data.shape)
data.head()

Part 2: Data Cleaning and Preprocessing¶

Before loading into Neo4j we need clean, unambiguous data:

Fill empty cast fields with empty strings (cast is optional)
Drop rows with missing director or country (required for graph edges)
Strip whitespace from title to avoid duplicate nodes

# Fill missing cast values.
data["cast"] = data["cast"].fillna("")
# Remove rows with missing director or country.
data = data.dropna(subset=["director", "country"])
# Strip whitespace from title to avoid duplicates.
data["title"] = data["title"].str.strip()
_LOG.info("Cleaned dataset shape: %s", data.shape)
data.head()

Part 3: Graph Construction with Neo4j¶

The Neo4jAPI wrapper class handles connection management and data loading. It uses MERGE instead of CREATE to prevent duplicate nodes:

MERGE (movie:Movie {title: $title, release_year: $release_year})
MERGE (director:Director {name: $director})
MERGE (director)-[:DIRECTED]->(movie)

This creates:

Movie nodes with title and release_year properties
Director nodes with a name property
[:DIRECTED] relationships from director to movie

# Start the Neo4j server.
!sudo neo4j start

# Initialize the Neo4j API wrapper.
neo4j_api = ttneouti.Neo4jAPI(
    uri="neo4j://localhost:7687",
    user="neo4j",
    password="new_password",
)

# Load the first 40 rows into Neo4j (lightweight demo).
neo4j_api.load_data(data[:40])
_LOG.info("Data loaded into Neo4j.")

Part 4: Graph Visualization¶

Query the database and render the Director → Movie graph using NetworkX:

MATCH (d:Director)-[r:DIRECTED]->(m:Movie)
WHERE d.name <> 'Unknown'
RETURN d.name AS director, m.title AS movie, m.release_year AS year

Visualization style:

Blue nodes: Directors
Green nodes: Movies (with release year)
Arrows: DIRECTED relationships
Layout: Spring layout for optimal spacing

# Generate the visualization.
neo4j_api.visualize_graph()

Clean Up¶

# Close the Neo4j connection.
neo4j_api.close()
_LOG.info("Connection closed.")