Data Analysis and Visualization of Bilibili Documentary Metadata

We first obtain the Bilibili documentary index page, inspect network requests, and identify the search API endpoint

https://bangumi.bilibili.com/media/web_api/search/result?...&page={3}&pagesize=20

and the season detail API

https://bangumi.bilibili.com/view/web_api/season?season_id={0}

.

Using these APIs we crawl basic information (title, year, season_id, media_id) into a list, convert it to a pandas DataFrame, and then request detailed fields such as danmakus, favorites, views, coins, area, episodes_duration, style, and episodes_aid for each season.

The collected data are saved to documentary_data_allinfo.csv and later re‑loaded for analysis.

Visualization steps include:

Region distribution via a pie chart of documentary counts per area.

Genre comparison for China, the UK, and the US using stacked bar charts and a co‑occurrence network graph.

Yearly documentary count trends across the three regions.

Episode length and season total duration distributions using histograms, violin plots, and 2‑D kernel density estimates.

Popularity analysis with a bubble chart that combines views, danmakus, coins, and favorites into a weighted score.

Comment‑time line plots for selected series, showing weekly comment dynamics.

Overall the analysis reveals that Chinese documentaries are dominated by history, society, and humanities topics with shorter episodes, while British and American productions feature more technology and nature content with longer episodes; popularity correlates strongly with coin and danmaku metrics.