Spotify Genre Classification

Term Project for Data Science II (Group 8)

📄 Click here to view the Full Project Report

A comparative analysis of six machine learning models for predicting the genre of songs on Spotify based on their audio features and metadata.

Group Members

• Anvita Yerramsetty
• Austin Bell
• Carter Prince
• Robera Abajobir
• Sanghyun An
• Tyler Varma

Project Overview

This project aims to classify music tracks into 24 distinct genres using 14 audio features extracted from the Spotify API. We implemented a complete data science pipeline including data cleaning, hybrid balancing, feature scaling, hyperparameter tuning, and comparative analysis.

The models evaluated are:

1. Logistic Regression
2. K-Nearest Neighbors (KNN)
3. Gaussian Naive Bayes
4. Random Forest Classifier
5. Gradient Boosting (XGBoost)
6. Multilayer Perceptron (Neural Network)

Directory Structure

.
├── data/                   # Contains processed CSV files and metadata
├── output/                 # JSON files containing results from each model run
├── preprocess.py           # Script to clean, balance, and scale the raw data
├── analyze.py              # Script to generate the results table and leaderboard
├── generate_figures.py     # Script to create all visualizations for the report
├── report.tex              # Final LaTeX report source code
├── report.pdf              # Final report PDF
├── requirements.txt        # Python dependencies
└── [model_scripts].py      # Individual training scripts (e.g., xgboost.ipynb, mlp.py)

How to Reproduce Results

1. Prerequisites

Ensure you have Python 3.8+ installed. Install the required dependencies:

pip install pandas numpy scikit-learn xgboost torch matplotlib seaborn

2. Data Preprocessing

The raw data (SpotifyFeatures.csv) is processed into training and testing sets. This script handles cleaning, encoding, scaling, and splitting.

python preprocess.py

Output: Generates X_train.csv, y_train.csv, X_test.csv, y_test.csv in the data/ folder.

3. Model Training

Each model has its own script or notebook. Running these will perform hyperparameter tuning and save the results to a standardized JSON file in the output/ directory.

Example:

python mlp.py
python gaussian_nb.py
# ... etc

Note: The output/ directory already contains the results from our final runs, so re-training is optional.

4. Analysis & Visualization

To generate the comparative leaderboard and the figures used in the report:

# Generates the leaderboard and efficiency plots
python analyze.py

# Generates specific figures (Class Balance, Correlation, Hyperparameter plots)
python generate_figures.py