And The Grammy Goes To...:
A Predictive Analysis of Grammy Award Outcomes
This is my senior thesis: a data-driven analysis of Grammy outcomes, using predictive modeling to uncover trends in music industry recognition.
Project Overview
What makes a song Grammy-worthy — commercial success, artistic merit, or some elusive balance of both?
This project investigates that question through a data-driven lens, using machine learning to predict Grammy Award winners from 2004 to 2025. I built a full pipeline for data collection, processing, modeling, and evaluation — drawing on audio features, Billboard chart performance, and Genius lyrics to analyze patterns across three major categories: Song of the Year, Record of the Year, and Best Rap Song.
​
The strongest predictive model combined broad trends with category-specific nuance, revealing that both lyrical richness and commercial momentum tend to correlate with Grammy wins — though exceptions underscore the subjectivity and politics of the award process. These findings shed light on the factors that drive recognition in the music industry and offer a new way to understand how cultural value is decided.
My Approach
To predict Grammy winners, I built a full machine learning pipeline — from data collection and cleaning to model training and evaluation.
​
I gathered and merged data from four key sources:
​​
​​
​
​
​
​
​
​
​
​
​​
​
​
​​​​​​
​
​​​
​
After cleaning and engineering features from this data, I trained XGBoost models to predict and rank nominees by their likelihood of winning. I explored both global models (trained across all categories) and category-specific models (tuned for each award).
​
The models were evaluated using a custom ranking metric to reflect real-world Grammy selection dynamics — and I even accounted for quirks like class imbalance and subjective snubs.
View the Full Presentation
Below are the complete slides from my thesis presentation, designed to visually walk through the project, from concept and pipeline to model performance and insights. Swipe through to explore the full story.
Full Thesis Paper
For a deep dive into the methodology, modeling choices, and key findings, you can read the full paper below. It includes detailed analysis, citations, and reflections on the broader implications of using machine learning in cultural recognition.