Bechdel Test Movie Analysis
Trends in gender representation across 9,000+ films, and ML models that predict whether a film passes the Bechdel Test.
- When
- Mar 2025 – Jun 2025
- Status
- Archived
- Links
- Code
- Python
- pandas
- scikit-learn
- seaborn
- networkx
Overview
Combined a Kaggle IMDb/Bechdel dataset (9,000+ films) with FiveThirtyEight’s dataset (~1,700 films with budgets and box office) to study gender representation in film over time.
Method
- Cleaning and merging on movie titles, resolving inconsistent formatting and missing values.
- Feature engineering: total gross, profit, release decade, and a pass/fail label.
- Visualization of trends by decade and genre with matplotlib, seaborn and networkx.
- Modeling: Logistic Regression, Decision Tree, Random Forest and KNN classifiers.
Results
Logistic Regression had high recall on passing films but struggled on failing ones; Decision Tree and Random Forest were more balanced. Restricting to 1970–2010 didn’t improve accuracy much but made the models easier to interpret.