Predicting movies’ eudaimonic and hedonic scores: A machine learning approach using metadata, audio and visual features

• DOI: 10.1016/j.ipm.2023.103610
• Publication Date: 2023-12-14T16:28:31Z
• Authors: Elham Motamedi; Danial Khosh Kholgh; Sorush Saghari; Mehdi Elahi; Francesco Barile; Marko Tkalcic
• Keywords: Baseline (sea); Predictive modelling; Feature (linguistics)

In the task of modeling user preferences for movie recommender systems, recent research has demonstrated benefits describing movies with their eudaimonic and hedonic scores (E H scores), which reflect depth message level fun experience they provide, respectively. So far, labeling E been done manually using a dedicated instrument (a questionnaire), is time-consuming. To address this issue, we propose an automatic approach predicting scores. Specifically, collected 709 from 370 users (with total 3699 records), augmented dataset metadata, audio, low-level high-level visual features, trained machine learning models movies. This study investigates use in various feature sets, including metadata. We compared performance predictive different combinations features majority classifier as baseline approach. The results demonstrate that our proposed learning-based significantly outperform scores, particularly when leveraging metadata features. random forest achieved 20% increase ROC AUC to both score score. These improvements were found be statistically significant. Overall, findings suggest automated tools are promising alternatives traditional questionnaire-based approaches.