Decoding Super Bowl Ad Popularity features
Year:
2023
Overview
This project, completed as part of DS5220 - Supervised Machine Learning, explores the factors driving the popularity of Super Bowl commercials. Using a dataset of 247 ads from 10 major brands, the study aims to uncover what makes an advertisement go viral - from humor and sex appeal to emotional or patriotic themes. Given that brands invest millions in these high-profile ads, understanding these drivers offers valuable insights into effective marketing and audience engagement strategies.
Approach
The dataset underwent extensive preprocessing, including feature selection, normalization, and missing value imputation to ensure high-quality model inputs. A custom target variable was engineered using multiple engagement thresholds (70%, 80%) based on likes, dislikes, and view counts.
Several machine learning models were tested, with a fine-tuned Random Forest classifier achieving the best performance in predicting ad popularity. The model’s interpretability helped identify the most influential features contributing to an ad’s success.
Key Features
Data-Driven Insights: Analyzed 247 Super Bowl ads from 10 leading brands to identify key success factors.
Feature Engineering: Designed custom engagement-based targets to quantify ad popularity.
Model Development: Trained and fine-tuned a Random Forest model for accurate prediction.
Interpretability: Extracted key predictors of ad success - animals, humor, and sex appeal.
Preprocessing Pipeline: Implemented end-to-end data cleaning and normalization workflows for consistent model performance.
