Review Aggregate Critical Film Success Predictor

Overview

Link: GitHub Repository

Using the Letterboxd review dataset, I implemented accuracy analysis metrics inspired by pollster ratings and outlier detection to identify reviewers with predictive power for critical success of films.

Key Methodologies

• Community Detection: Formed aggregates of reviewers via community detection algorithms.
• Machine Learning & Time-Series: Used ML models alongside time-series analysis to understand reviewer sentiment trends.
• Alpha Generation: Generated predictive signals for award outcomes, analogous to alpha generation, with planned applications including finding mispricing on financial betting markets.

Organization

• The Erdős Institute