IMDb Movie Rating Prediction Using a Random Forest Classification Approach

Abstract

Accurate movie rating prediction is essential for supporting audience preferences and analytical decision-making in the digital film industry. The availability of large-scale metadata from IMDb provides valuable opportunities for applying machine learning techniques to analyze rating patterns. This study investigates the effectiveness of a Random Forest classification model for predicting IMDb movie rating categories based on structured attributes, including genre, movie duration, content rating, actor popularity, and user review statistics. Data preprocessing involved handling missing values, removing duplicates, encoding categorical variables, normalizing numerical features, and partitioning the dataset into training and testing subsets. To mitigate class imbalance among rating categories, the Synthetic Minority Over-sampling Technique (SMOTE) was applied to the training data. Experimental evaluation demonstrates that the proposed model achieves an overall accuracy of 0.78, accompanied by balanced precision, recall, and F1-score values across all classes. Confusion matrix analysis shows that classification errors predominantly occur between neighboring rating categories, reflecting the inherent subjectivity of movie ratings. Furthermore, feature importance analysis highlights genre, duration, content rating, and user engagement indicators as the most influential predictors. These results indicate that Random Forest offers a robust and interpretable baseline model for IMDb rating prediction and provides meaningful insights for future movie analytics and recommendation research.