Predicting Video Popularity of the Military Museum: A TikTok Case Study

This study proposes an innovative deep learning model that combines Convolutional Neural Networks (CNN), Convolutional Block Attention Module (CBAM), and Transformer for the precise prediction of short video popularity. Through comparative experiments with several models, including DNN, RNN, CNN, CNN + Transformer, and CNN + CBAM + Transformer, the results show that the CNN + CBAM + Transformer model significantly outperforms others in classification accuracy (96.73%) and other performance metrics. The model utilizes CNN to extract local features of the video, enhances focus on key features through CBAM, and captures global dependencies and complex patterns using Transformer, thereby improving the overall predictive capability. Compared with traditional models, this method demonstrates greater adaptability and robustness in handling the multidimensional features of video popularity prediction. By accurately predicting video popularity, this study not only helps platforms optimize recommendation algorithms and enhance user experience, but also provides high-value data support for content creators and advertisers. The innovative model proposed in this paper offers a new paradigm for short video popularity prediction, with significant theoretical and practical implications.