Country by Country Comparison of Thumbnail Features Contributing to Views Using AIME for YouTube.

This study proposes a thumbnail feature extraction method that contributes to the number of views on YouTube using AIME and examines the differences in thumbnail features by country using this method. In recent years, video media content, including YouTube, has become scattered all over the Internet and continues to increase enormously. For users to click and play this video media content, it is important to present appropriate thumbnails to creators. However, it is not known what kinds of thumbnail features contribute to the number of views. In this study, we propose a new method to extract thumbnail features that contribute to the number of views using AIME, which can extract features that contribute to the target variable by deriving the approximate inverse operator of a machine learning black box model. Furthermore, we collected YouTube videos from the U.S., Canada, Japan, France, Germany, the U.K., and Australia to show the differences in thumbnail features.