Identifying content unaware features influencing popularity of videos on YouTube: A study based on seven regions. (15th November 2022)

• Record Type: Journal Article
• Authors: Halim, Zahid; Hussain, Sajjad; Hashim Ali, Raja
• Is Part Of: Expert systems with applications. Volume 206(2022)
• Publication Date: 2022-11-15
• Subjects: Video classification -- YouTube -- Artificial intelligence -- Content popularity -- YouTube trend prediction; Expert systems (Computer science) -- Periodicals; Systèmes experts (Informatique) -- Périodiques; Electronic journals; 006.33
• DOI: 10.1016/j.eswa.2022.117836
• Languages: English
• ISSNs: 0957-4174

Abstract

Highlights: AI techniques to identify the satellite features that make YouTube video trending. AI-based methodology that assesses the impact of various content-agnostic factors. Dataset is created from YouTube for seven regions. Feature selection techniques executed on to extract important attributes. Helpful for content creators and YouTubers to make their video trending. Abstract: Predicting the popularity of User Generated Content (UGC) is a subject of interest to the Internet service providers, content makers, social media researchers, and online advertisers. However, it is also a challenging task due to multiple factors that influence social networks' content popularity. This work utilizes the Artificial Intelligence (AI) techniques to identify the features that contribute towards a video to enter into the trending category on YouTube. It examines the data generated by a video and its potential to get trending. For this, the present work utilizes AI for feature prediction. An AI-based methodology is presented that assesses the impact of various content-agnostic factors regarding video popularity in seven different regions, including Canada, France, Germany, India, Pakistan, United Arab Emirates, and the United States of America. A dataset is extracted from YouTube for these regions, and feature selection techniques are executed on the datasets to extract important attributes. A class label is assigned to each video, and the dataset is profiled having one of the two class labels, i.e., trending or non-trending. The top three features for each video (region wise) are obtained. It is observed that the trending behavior is dissimilar in different regions. Finally, three classifiers, namely, artificial neural networks, k- Nearest Neighbor, and support vector machine, are trained to predict if a video can get into the trending category on YouTube. The proposed solution is compared with two closely related state-of-the-art methods. This work is useful for content creators and YouTubers to make their video trending and more appealing.