This project analyzes your YouTube watch history using data from your Google Takeout export. It is divided into two sections:

  1. Basic Analysis (No API) — using only the raw JSON data
  2. Extended Analysis (with YouTube API) — enriched with additional video metadata (e.g., duration, category, tags)

Section One: JSON-Only Analysis

The script processes your watch history and filters out:

  • Shorts and other short-form content
  • Ads
  • Videos watched within 1 minute of each other (to reduce noise)
  • Deleted or private videos
  • Music-related content (based on keywords in channel names)
  • Videos with only hashtags or URLs

After filtering, the data is converted into a pandas DataFrame with the following columns:

  • title
  • channel
  • datetime
  • video_id

Outputs & Visualizations

  • 📌 Total and unique video counts
  • 🔁 Most frequently rewatched videos
  • ☁️ Word cloud based on video titles
  • 🗓️ Videos watched per year, quarter, and month
  • ⏰ Viewing patterns by hour, weekday, and month
  • 📺 Top channels overall and per quarter
  • 🧠 AI-generated insights into your viewing habits
  • 🎯 Personalized video suggestions for future learning

Section Two: API-Enhanced Metadata (WIP)

This part will use the YouTube Data API to fetch:

  • Video duration
  • Category
  • Tags
  • Region-specific metadata

Currently a work in progress.

GitHub repo

Rating: 4/5

Back to projects