Social Network Analysis of Netflix Collaborations 🎬

Analyzing Collaboration Patterns Between Directors & Actors Across Movie Genres
A Master's Course Project in Social Network Analysis (SNA)

---

📌 Overview

This project explores collaboration networks in the U.S. film industry using Netflix data. By applying Social Network Analysis (SNA), we uncover how directors and actors collaborate within and across genres, identify key influencers, and track how these networks evolve over time. Key questions answered:

• Homophily: Do creators prefer working within the same genre?
• Centrality: Who are the most influential directors/actors in each genre?
• Dynamics: How do collaborations change over time?
• Communities: Do genres form distinct collaboration clusters?

---

📂 Dataset

• Source: Netflix Movies and TV Shows (Kaggle)
• Scope: Focused on 2,684 U.S.-produced movies (filtered from the original dataset).
• Key Metadata:
  • Directors, actors, genres, release year.
  • Genres analyzed: Drama, Comedy, Documentary, Horror, Children & Family, Action & Adventure.

---

🛠️ Methodology

1. Graph Construction

• Nodes: Directors (_d suffix) and Actors (_a suffix).
• Edges: Created if two individuals collaborated on the same movie.
• Genre-Specific Networks: Separate graphs built for each genre.

2. Key Metrics

Metric	Definition	Relevance
Degree Centrality	Number of connections a node has.	Identifies highly collaborative individuals.
Betweenness Centrality	How often a node acts as a bridge between others.	Highlights "connectors" in the network.
Closeness Centrality	Average distance from a node to all others.	Measures how quickly someone can reach others.
Clustering Coefficient	Likelihood that two collaborators of a node are also connected.	Indicates tight-knit groups (e.g., frequent repeat collaborations).
Preferential Attachment	Probability of new connections forming with highly connected nodes ("rich get richer").	Explains growth patterns in networks.

3. Temporal Analysis

• Cumulative Networks: Built yearly from 1990–2021 to track evolution.
• Metrics tracked over time: Average degree, distance, clustering, preferential attachment.

4. Community Detection

• Algorithms: Multi-level and Leiden (resolution-based).
• Modularity: Measures how well a network is divided into communities.

---

🔍 Key Findings

1. Genre-Specific Networks

• Horror Films: Tight-knit communities (high clustering). Directors act as central hubs.
• Documentaries: Fewer connections per person but many small clusters.
• Comedies: Actors have the highest closeness centrality (well-connected within the genre).

2. Central Figures

• Directors: Martin Scorsese (Drama), Scott Stewart (Horror).
• Actors: Samuel L. Jackson (Action), James Franco (Drama).

3. Temporal Trends

• Actors in Dramas/Comedies: Exponential growth in collaborations post-2000.
• Documentaries: Surge in director involvement (2015–2020).
• Shrinking Distances: Actors became more interconnected over time.

4. Communities Reflect Genres

• Comedy and Children & Family actors form distinct clusters.
• Mixed-genre communities (e.g., Drama + Action) also exist due to multi-genre collaborations.

---

💻 Code & Reproducibility

Dependencies

• Python 3.8+
• Libraries: pandas, networkx, matplotlib, seaborn.

How to Run

1. Data Preparation:

   df = pd.read_csv('https://raw.githubusercontent.com/aphdinh/socialnetwork/main/netflix_titles.csv')  
   # Filter U.S. movies and preprocess (see notebook for details)  

Build Graphs:

G = create_graph_from_df(usa)  # Custom function to generate collaboration networks  

Analyze Metrics:

compute_average_degree(G, "actor", "Comedy")  # Example: Avg. degree for Comedy actors  

Visualize Results:

• Pre-built functions for plotting trends (e.g., plot_metric()).

📈 Key Takeaways for Practitioners

• Network Insights: Genre-specific collaboration patterns can inform talent recruitment or partnership strategies.
• Temporal Dynamics: Recent genres (e.g., Horror) show denser networks, indicating active communities.
• Toolkit: Code provides reusable functions for SNA on collaboration datasets.

📚 References

• Concepts: Preferential Attachment (Barabási–Albert Model), Centrality Measures.
• Data Source: Kaggle Netflix Dataset.