The Impact of Explainable AI on Low-Accuracy Models: A Practical Approach with Movie Genre Prediction

This study explores the challenges of classifying film genres using an IMDB dataset for in-depth analysis. Despite the inherent difficulties of genre overlap, the study refines genres into eight core categories. The best performing XGBoost model, using TF-IDF vectorization and metadata, achieves an accuracy of 74% after optimization. Notably, the study places special emphasis on Explanable Artificial Intelligence and introduces LIME to provide insights into genre categorization beyond traditional accuracy metrics. It shows that despite low comparative accuracy, the classifications are humanly understandable. The importance of XAI in genre classification is further emphasized, arguing for a deeper evaluation beyond traditional accuracy metrics. This research not only contributes to the ongoing refinement of genre classification models, but also serves as an effort to incorporate XAI to improve the understanding and interpretation of trained models.