Video Popularity in Social Media: Impact of Emotions, Raw Features and Viewer Comments

TL;DR

This study investigates how emotional reactions and viewer comments influence the popularity of environmental videos on social media, finding comments and reactions are more impactful than raw video features.

Contribution

It introduces a comparative analysis of raw video features versus viewer engagement metrics in predicting social media video popularity.

Findings

Viewer comments and reactions significantly influence popularity (accuracy 0.8).

Raw video features have a moderate influence (accuracy 0.67).

Sadness in comments correlates positively with likes, negative sentiment correlates negatively.

Abstract

The Internet has significantly affected the increase of social media users. Nowadays, informative content is presented along with entertainment on the web. Highlighting environmental issues on social networks is crucial, given their significance as major global problems. This study examines the popularity determinants for short environmental videos on social media, focusing on the comparative influence of raw video features and viewer engagement metrics. We collected a dataset of videos along with associated popularity metrics such as likes, views, shares, and comments per day. We also extracted video characteristics, including duration, text post length, emotional and sentiment analysis using the VADER and text2emotion models, and color palette brightness. Our analysis consisted of two main experiments: one evaluating the correlation between raw video features and popularity metrics and another assessing the impact of viewer comments and their sentiments and emotions on video popularity. We employed a ridge regression classifier with standard scaling to predict the popularity, categorizing videos as popular or not based on the median views and likes per day. The findings reveal that viewer comments and reactions (accuracy of 0.8) have a more substantial influence on video popularity compared to raw video features (accuracy of 0.67). Significant correlations include a positive relationship between the emotion of sadness in posts and the number of likes and negative correlations between sentiment scores, and both likes and shares. This research highlights the complex relationship between content features and public perception in shaping the popularity of environmental messages on social media.