🎵 Spotify Mood Ring

Does popular music get happier or sadder over time — and does it differ by genre?

This project scrapes Billboard's Year-End Hot 100 chart data across multiple years, joins it with public audio-feature data (valence, energy, tempo, danceability), and analyzes how musical "mood" shifts year-over-year and across genres.

Built as a hands-on project to practice: Python fundamentals, NumPy vectorized operations, Pandas DataFrames, boolean/filtering logic, web scraping, and reshaping/pivoting data.

---

🎯 Project Questions

• How does average song "valence" (musical positivity) trend year-over-year within each genre?
• Which genres show the widest mood swings across different years?
• Are there outlier year-genre combinations that break the typical trend (e.g. an unusually happy or somber year for a genre)?
• Is there a correlation between a song's tempo/energy/danceability and its valence?

Note: this project originally aimed at weekly seasonal analysis, but pivoted to annual Billboard Year-End data after determining that live weekly Spotify chart scraping wasn't feasible (see Data Sources below). The questions above reflect what the final annual dataset supports.

Findings

1. Valence trend by genre: British valence decrease from 0.25 in 2001 to 2022, while country stayed relatively flat across the same period.
2. Widest mood swings: synth-pop had the widest year-to-year valence swings (std ≈ 0.333).
3. Outliers: 85 year-genre combinations were flagged as statistical outliers. The most extreme was indie in 2011 (valence 0.965) — a notably upbeat year compared to that genre's typical range.
4. Correlations: Valence shows a moderate positive correlation with danceability (r=0.34) — happier songs tend to be more danceable — a weaker positive link with energy (r=0.21), and a moderate negative correlation with tempo (r=-0.31), suggesting faster songs trend slightly less positive.

---

📁 Repository Structure

spotify-mood-ring/
├── data/
│   ├── raw/                 # Untouched scraped HTML/JSON and downloaded audio-feature CSVs
│   └── processed/           # Cleaned, merged, analysis-ready DataFrames (parquet/csv)
│
├── notebooks/               # Exploratory analysis, step-by-step, one notebook per project phase
│   ├── 01_scrape_explore.ipynb
│   ├── 02_clean_merge.ipynb
│   ├── 03_filtering_questions.ipynb
│   ├── 04_pivot_seasonal_mood.ipynb
│   └── 05_numpy_stats_plots.ipynb
│
├── src/                      # Reusable code, imported by notebooks/scripts (not copy-pasted)
│   ├── __init__.py
│   ├── scraper.py            # Chart-scraping functions (parse_chart_page, retry logic, etc.)
│   ├── clean.py              # Normalization/cleaning functions (dedupe, name matching)
│   └── analysis.py           # Pivoting, rolling stats, z-scores, correlation helpers
│
├── tests/                    # Unit tests for src/ functions
│   ├── __init__.py
│   └── test_clean.py
│
├── outputs/                  # Saved charts/plots (PNG) for the README or portfolio writeup
│
├── .gitignore
├── requirements.txt
└── README.md

---

🛠️ Setup

git clone <your-repo-url>
cd spotify-mood-ring
python -m venv venv
source venv/bin/activate       # Windows: venv\Scripts\activate
pip install -r requirements.txt

---

🚀 Usage

1. Scrape chart data (fills data/raw/)

   python -m src.scraper

2. Clean & merge with audio features (fills data/processed/)

   python -m src.clean

3. Run the analysis (computes rolling trends, outliers, correlations, and saves the chart)

   python -m src.analysis

4. Or explore step-by-step in notebooks, in order, starting with notebooks/01_scrape_explore.ipynb

---

📊 Skills Practiced

Skill	Where
Python fundamentals (functions, control flow)	src/scraper.py, src/clean.py
Web scraping / data extraction	src/scraper.py, notebooks/01
Pandas Series/DataFrame basics	notebooks/02, src/clean.py
Filtering & boolean indexing	notebooks/03
Reshaping / pivoting (pivot_table, melt)	notebooks/04, src/analysis.py
NumPy vectorized ops (rolling avg, z-scores, correlation)	notebooks/05, src/analysis.py

---

📈 Sample Finding

Acoustic tracks showed the widest year-to-year valence swings (z-scores over 1.0 in multiple years), while alt-rock stayed comparatively stable across the same period. Valence correlates moderately with danceability (r=0.33) but has a weak negative relationship with tempo (r=-0.32), suggesting faster songs trend slightly less "positive" on average.

---

📄 Data Sources

• Chart data: Wikipedia's "Billboard Year-End Hot 100 singles" pages (one static, server-rendered page per year — see below for why the original weekly Spotify-chart scraping idea wasn't achievable with requests/BeautifulSoup, since Spotify's charts are loaded client-side via a private API)
• Audio features: a public Kaggle Spotify Tracks Dataset | Audio Features dataset, downloaded once and joined to the scraped chart data by normalized song title + artist name

---

🎧 Fun Fact

I love music — especially the phonk genre. My very first ever project was actually an open-source music player called Phonq. This project brings that same love of music into a data science lens.

---

🔭 Possible Extensions

• Add sentiment analysis on lyrics (would introduce basic NLP)
• Pull monthly/weekly chart data from a source that supports it, to revisit the original seasonal-mood question
• Add a chart-longevity metric (e.g. weeks on chart) to test tempo/energy vs. longevity
• Build a small dashboard (Streamlit) to explore mood-by-genre interactively