The Effect of Keywords on the Number of Likes, Comments, and Views in Douyin Videos

Based on massive user behavior data from the Douyin platform, this study employs Python-based data analysis techniques to conduct in-depth mining and visualization of key video performance metrics. Through a systematic analysis of core indicators such as the number of likes, comments, and views, combined with user verification status and activity characteristics, the key factors influencing video interaction effectiveness are revealed. The research findings include: 1) Video performance exhibits a typical long-tail distribution, with 80% of videos concentrated in the low like count range (100,000) demonstrate distinct content characteristics, with videos containing keywords such as "expose" or "tutorial" in their titles exhibiting an interaction rate 47% higher than the average; 4) User activity level is positively correlated with content production quality, with active users contributing 78% of high-quality content. Innovatively utilizing TF-IDF keyword extraction and regression modeling, the study identifies the ten most influential content tags, such as "technology" and "life tips." Through multi-dimensional visualizations including heatmaps and scatter plots, it provides data-driven support for content creators to optimize topic selection and for platforms to enhance recommendation algorithms. The constructed interactive analysis dashboard significantly improves decision-making efficiency. These findings offer valuable insights for the refined operation of the short video industry.