Research on Movie Box Office Prediction Model Based on Machine Learning

With the unprecedented success of "Ne Zha 2" in 2025, China's animation film industry has achieved a historical breakthrough admist the global cinema markets exponential growth. As box office prediction emerges as an essential tool for investment optimization and creative development, this study establishes a machine learning-based prediction framework analyzing core determinants of commercial performance. Utilizing the Kaggle movie dataset with USD-standardized financial metrics, this paper integrates multivariate features such as budget, genre, director, and audience ratings. Through comparative analysis of ridge regression, random forest, and XGBoost models enhanced by nested cross-validation and Bayesian hyperparameter optimization, the refined random forest model demonstrated superior predictive capability (RMSE = 8.610, R = 0.89), effectively capturing non-linear interactions. Feature importance analysis indicates that the number of audience votes (vote_count) and budget account for over 60% of the influence on box office performance, while movie runtime has relatively limited impact. The frameworks practical applications span three strategic dimensions: from optimizing fund allocation, adjusting creative directions, to scheduling strategies. This research advances film analytics methodology by establishing an interpretable prediction paradigm, while proposing future extensions incorporating real-time social media sentiment analysis and cross-cultural adaptation models, holding significant theoretical and practical value.