Movie Success Forecasting and Audience Clustering Using XGBoost and Fuzzy C-Means

In the rapidly evolving film industry, predicting movie popularity and identifying the target audience during the pre-production phase is crucial for producers, marketers, and distributors. This paper presents a dual-objective framework aimed at forecasting a movie's potential success and segmenting its target audience using machine learning techniques. The system employs a content-based recommendation approach that analyzes key pre-production features-such as genre, cast, director, and plot-to identify similar movies from an IMDb-based dataset. These similar movies serve as the foundation for a supervised learning model using XGBoost, which predicts the likely success of upcoming films. XGBoost demonstrated superior performance compared to CNN and Random Forest in terms of accuracy, precision, recall, and F1-score. The model classifies movies into four categories: Super Hit, Hit, Average, and Flop, based on predicted popularity scores. For audience segmentation, the system uses the Fuzzy C- Means (FCM) clustering algorithm, which outperforms classical methods like K - Means and Gaussian Mixture Models (GMM) by enabling soft membership across overlapping age preferences. The audience is grouped into four predefined categories: Junior (below 13), Teenage (13–19), Mid-age (20–50), and Senior (above 50), based on industry-standard demographic brackets. While the training phase includes ratings and votes to improve the model, the prediction for unreleased movies relies solely on pre-release metadata, making it suitable for early-stage decision-making. This hybrid framework offers a data-driven solution that enhances decision-making in film production and marketing by minimizing investment risk and aligning promotional strategies with audience insights.