🎵 The Sound of Union 🎵

Quantifying Union College’s Music Preferences with Logistic Regression

Author: Conor Fryer
Course: STA 264 – Hoerl – Winter 2025

Overview

This project explores whether Union College students have distinct musical preferences that can be quantified using logistic regression. Built on Union Rewind, an annual WRUC initiative, the study compares students' most-played songs to a dataset of globally popular tracks to determine if certain song characteristics influence inclusion in a Union student’s top five.

Using Spotify’s audio features, a logistic regression model was developed to assess measurable predictors like danceability, energy, acousticness, and instrumentalness. The results revealed significant trends—most notably, highly danceable music, despite being dominant in social settings, was less likely to appear in students' personal top five lists.

---

🔍 Key Features:

• Data Collection: Union Rewind survey data (students' self-reported top five songs) vs. a dataset of globally popular songs.
• Statistical Analysis: Logistic regression applied to Spotify audio features.
• Insights: Identifying measurable trends in Union students' music preferences.
• Tech Stack: JMP, Python (for data processing), Spotify API (for feature extraction).

---

📂 Repository Structure

unionrewindregressionanalysisproject/
│-- data/               # Contains raw and processed datasets
│-- jmp/                # JMP analysis files
│-- src/                # Python scripts (Spotify API extraction, data preprocessing)
│-- LICENSE             # MIT License
│-- README.md           # Project documentation
│-- .gitignore          # Specifies ignored files (e.g., .env)

---

⚙️ Configuration

This project requires Spotify API credentials for extracting song features. You must create a .env file in the root directory with the following:

SPOTIFY_CLIENT_ID=your_client_id_here
SPOTIFY_CLIENT_SECRET=your_client_secret_here
SPOTIFY_REDIRECT_URI=your_redirect_uri_here

🔧 Steps to Set Up:

1. Register your application on the Spotify Developer Dashboard to obtain API credentials.
2. Set the redirect URI in your Spotify Developer settings.
3. Create a .env file in the project root and add the credentials.

🔹 Pro Tip: To prevent accidental commits, the .env file is ignored by Git. Instead, you can use the provided .env.example template:

# Rename this file to .env and fill in your credentials
SPOTIFY_CLIENT_ID=your_client_id_here
SPOTIFY_CLIENT_SECRET=your_client_secret_here
SPOTIFY_REDIRECT_URI=your_redirect_uri_here

---

Data Collection

• Included Dataset: Collected through Union Rewind, where Union College students reported their top five songs via an in-person survey.
• Non-Included Dataset: Drawn from Kaggle's "Top Spotify Songs in 73 Countries" dataset, representing globally popular music.

---

Methodology

• Logistic Regression Analysis:
  • Predicts whether a song was included in a Union student's top five based on Spotify audio features.
  • Key predictors analyzed: danceability, energy, acousticness, instrumentalness, speechiness, valence, and live elements.
  • Popularity and loudness were removed due to their overwhelming influence and redundancy.
• Model Validation:
  • Alternative datasets were tested to confirm consistency.
  • Statistical tests ensured robustness.

---

Key Findings

• Distinct Preferences: Union students’ musical choices differed from global streaming trends.
• Danceability: Strong negative effect—highly danceable songs were less likely to appear in personal top-five selections.
• Instrumentalness: Only feature with a positive effect—songs with instrumental elements had a slightly higher chance of inclusion.
• Energy & Speechiness: Moderately negative effects, indicating a preference for structured, polished music over raw or spoken-word-heavy tracks.
• Model Limitations: Music preference remains highly subjective, making it difficult to fully capture personal taste with statistical modeling.

---

Tools & Technologies

• Spotify API & Chosic Playlist Analyzer (for extracting audio features)
• JMP Statistical Software (for logistic regression analysis)
• Python & Pandas (for data processing and structuring)

---

🚀 Future Improvements

• Expanding analysis to multi-year trends in Union College’s music preferences.
• Exploring additional Spotify metadata (e.g., genre classifications) to refine preference modeling.
• Implementing interactive visualizations to display trends dynamically.

---

📜 License

This project is licensed under the MIT License, allowing free use, modification, and distribution with attribution.

Important: If you intend to redistribute Spotify data, please review Spotify’s Developer Terms to ensure compliance.

---

🏆 Acknowledgments

• Union Rewind participants for contributing their top songs.
• WRUC (Union College Radio Club) for facilitating data collection.
• Spotify API for providing structured audio feature data.