Youtube Comment Sentiment and Video Popularity¶

Motivation¶

Problem¶

For many popular Youtube creators, making videos is their main source of income. Creators make money via advertisements and sponsorships in their videos. The more reach (views) they have, the more money they garner.

Moreover, it would very useful for creators to know how their audience is receiving their content and if that is translating to a more popular video (i.e., higher income).

Solution¶

Youtube is the most popular platform for video creation, which has resulted in a multitude of public datasets on video statistics. The goal of this project is to construct a relationship between video statistics and audience sentiment. We could denote the video statistics as a measure of popularity and the audience sentiment could be measured via comment sentiment.

Impact¶

This project could prove extremely useful for Youtube creators as a demonstrated relationship between video popularity and audience sentiment could shed light on the reception of their content.

Data¶

Kaggle Youtube Statistics

Features (video_stats)¶

  1. Title: title of video
  2. Video ID: unique video identifier
  3. Published At: date of publication
  4. Keyword: genre of video
  5. Likes: number of likes video has
  6. Comments: number of comments video has
  7. Views: number of views video has

Features (comments)¶

  1. Video ID: unique video identifier
  2. Comment: audience feedback, a video comment
  3. Likes: number of likes a comment has
  4. Sentiment: sentiment score of a comment, sentiment > 0 yields a positive sentiment, sentiment < 0 yields a negative sentiment, and 0 neutral.

We have a variety of video statistics, which can be used to assess video popularity. In addition, we have comments associated with each video and their sentiment score, which can be used to assess audience sentiment.

In [1]:
import pandas as pd

video_stats = pd.read_csv('data/videos-stats.csv')
comments = pd.read_csv('data/comments.csv')

video_stats.head()
Out[1]:
Unnamed: 0 Title Video ID Published At Keyword Likes Comments Views
0 0 Apple Pay Is Killing the Physical Wallet After... wAZZ-UWGVHI 2022-08-23 tech 3407.0 672.0 135612.0
1 1 The most EXPENSIVE thing I own. b3x28s61q3c 2022-08-24 tech 76779.0 4306.0 1758063.0
2 2 My New House Gaming Setup is SICK! 4mgePWWCAmA 2022-08-23 tech 63825.0 3338.0 1564007.0
3 3 Petrol Vs Liquid Nitrogen | Freezing Experimen... kXiYSI7H2b0 2022-08-23 tech 71566.0 1426.0 922918.0
4 4 Best Back to School Tech 2022! ErMwWXQxHp0 2022-08-08 tech 96513.0 5155.0 1855644.0
In [2]:
comments.head()
Out[2]:
Unnamed: 0 Video ID Comment Likes Sentiment
0 0 wAZZ-UWGVHI Let's not forget that Apple Pay in 2014 requir... 95.0 1.0
1 1 wAZZ-UWGVHI Here in NZ 50% of retailers don’t even have co... 19.0 0.0
2 2 wAZZ-UWGVHI I will forever acknowledge this channel with t... 161.0 2.0
3 3 wAZZ-UWGVHI Whenever I go to a place that doesn’t take App... 8.0 0.0
4 4 wAZZ-UWGVHI Apple Pay is so convenient, secure, and easy t... 34.0 2.0

Methods¶

We will use regression, seeking to estimate video popularity via audience sentiment. Doing so will allow us to see any relationship between audience sentiment and video popularity.