🎬 Netflix Pandas Mini-Project

Week 2 — Joining & Combining Data with Pandas

(Single-Source Dataset With Generated Auxiliary Tables)

This project demonstrates core data-joining techniques in Pandas using only a single source file: netflix_titles.csv.
To simulate real multi-table data engineering workflows, we generate two additional datasets inside the notebook:

• credits — extracted from director and cast columns
• ratings — synthetic user rating events for each title

This allows us to practice a wide range of Pandas merge operations without needing external files.

---

📂 Project Structure

netflix-pandas-project/
│
├── data/
│ └── netflix_titles.csv # original dataset
│
├── notebook/
│ └── netflix_pandas_project.ipynb # full project notebook
│
├── README.md

---

📘 Objectives

This project focuses on mastering:

✔ merge() — one-to-one & one-to-many joins

✔ concat() — vertical stacking

✔ merge_asof() — time-aware joins

✔ merge_ordered() — ordered merges

✔ synthetic feature creation

✔ aggregation after merging

These skills are fundamental for real-world data engineering, EDA, and ML pipelines, where datasets rarely live in one place.

---

🛠️ Steps Performed

1. Load & inspect netflix_titles.csv

• Reviewed structure
• Parsed dates
• Cleaned key columns

2. Generated a synthetic credits table

Derived from:

• show_id
• title
• director
• cast

Rows with no director and no cast were removed to simulate missing metadata.

3. Generated a synthetic ratings table

Created using random sampling:

• 1–5 star user rating scores
• rating counts
• random rating_date between 2015–2021

One title may appear multiple times (one-to-many relationship).

4. Performed joins

Examples:

• Left join: titles + credits
• One-to-many join: titles + ratings
• Aggregation: mean rating per title
• Concatenation: movies + TV shows
• Time-based merge: ratings + marketing signals

---

📊 Key Insights

• Joins with missing values illustrate real challenges in metadata pipelines.
• One-to-many joins require aggregation to return to a single-row-per-title structure.
• merge_asof() efficiently combines time-series data without exact matches.
• concat() is essential for dataset stacking and feature-level engineering.

---

💡 Why This Project Matters

In real analytics and ML work, most datasets come in multiple tables, and your ability to cleanly join and transform them determines:

• data quality
• model accuracy
• feature engineering efficiency
• project maintainability

This project builds the foundation for:

✔ Feature engineering in Week 3
✔ ML preprocessing (Week 4–5)
✔ End-to-end ML pipelines (Week 6–12)

---

🚀 Next Steps

After this project, you are prepared to:

• Aggregate large datasets
• Build ML-ready feature tables
• Implement time-series merges
• Clean and combine multi-source datasets

These skills directly connect to Week 3 EDA, Week 4 Supervised ML, and Week 5–6 Pipelines & Feature Engineering.

---

📝 Author

Nicolas
Data Science & Machine Learning Engineer in training