nirmal-25/youtube-analytics-and-sentiment-analysis

Interactive data visualizations for trending YouTube videos using Sentiment Analysis in D3.js and Tableau

YouTube Trending Video Analytics

The repo consists of visualizations in D3.js (Observable) and Tableau for videos from YouTube's API for trending videos in the US. Refer to this repository to scrape data using the public API. The visualizations are created using raw statistics from the dataset, along with derived attributes like sentiment scores to gain more insight.

Motivation

• Visualizations on trending YouTube video data that help narrow down key indicators that makes a video popular, and helps stay on YouTube's trending list.

• What makes trending YouTube videos more engaging than others? How can content creators better make use of data to create high-performing videos that go viral?

Dataset

Download the CSV file for the dataset from here.

This is a cleaned version of the scraped data, and NLTK's SentimentAnalyzer is used to derive sentiment scores (ranging from -1 to 1) for videos, which are added to the CSV file as an additional feature.

Each entry in the dataset corresponds to trending videos (2020 to 2022) with features such as the title, trending date, published date, channel title, category, tags, description, views, likes, dislikes, comments, and comment count.

Visualizations

• Net popularity vs. sentiment score
• Category based interactive visualization
• Category-wise analysis of sentiment scores
• Trend analysis of sentiment scores
• Correlation for raw video statistics ...

For more details on these visualizations and helper scripts, go to misc/

Follow the steps given below to run the Observable notebook.

The Tableau visualizations can be found here.

Screenshots for sample visualizations are shown below.

Trend analysis of sentiment scores

Trend analysis of sentiment scores

Sentiment category distribution for top-n categories based on popularity

Average sentiment scores trend (per day)

Observable

To run the Observable notebook, upload the CSV file to the notebook

https://observablehq.com/d/dbd2e89e6a968b57@790

Alternatively, clone the repo and view this notebook in your browser by running a web server in this folder. For example:

npx http-server

Or, use the Observable Runtime to import this module directly into your application. To npm install:

npm install @observablehq/runtime@5
npm install https://api.observablehq.com/d/[email protected]?v=3

Then, import your notebook and the runtime as:

import {Runtime, Inspector} from "@observablehq/runtime";
import define from "dbd2e89e6a968b57";