AAI_2025_Capstone_Chronicles_Combined

Cinema Analytics and Prediction System

30

The overall classification performance comparison (Figure 31) shows that the binary

XGBoost classifier, which distinguishes between Flop and Hit, performed the best. A stratified

Dummy classifier was used as a baseline, producing predictions based on class distribution and

scoring around 0.5 across all metrics.

Figure 31: Classification model comparison

To conclude, our system demonstrated strong performance in analyzing and predicting

movie-related outcomes using both predictive modeling and NLP approaches. While results

were encouraging, real-world factors such as outliers and class overlap presented challenges.

The combined insights provide a foundation for building a comprehensive analytics platform,

with future efforts aimed at real-time deployment, continuous improvement, and interactive

tools to support data-driven decisions.

197

Internal