International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 09 Issue: 05 | May 2022
p-ISSN: 2395-0072
www.irjet.net
YouTube Trending Video Dashboard Kaustubh1 1UG
Student, Maharaja Agrasen Institute of Technology, Dept. of IT, Delhi, India ---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - YouTube is not only serving as an entertainment
a core set of variables and numerous categorization algorithms, researchers first anticipate the future popularity levels of videos. The study then used sophisticated regression models to forecast the number of views based on the popularity levels. (Ouyang, Li and Li, 2016) [1].
platform for the films and the television industry but it has also emerged as a learning platform for many students. Content creators on YouTube, commonly known as “YouTubers” are pushing their content every single day and hour to be relevant to their audiences. It’s a known fact that the YouTube algorithm is not a publicly available code and it is kept private for most of the time. Also, YouTube has shared any intuition about what factors are considered for a video to be in trending section which leaves its audience in dilemma about posting videos. The trending section is where some videos are listed as trending and it is designed in such a way that every user of this platform checks it once in a while. This dashboard aims to present the trending section of YouTube for India in terms of the trends and observations which could be helpful for creators to push their content to more audiences.
The impact of meta-data elements such as title, tag, thumbnail, and description on the popularity and trendiness of YouTube videos was investigated by Hoiles, Aprem, and Krishnamurthy (Hoiles, Aprem and Krishnamurthy, 2017).[2] Their study used a variety of Machine Learning algorithms to predict the YouTube video popularity based on the video's meta-features as well as other factors like the number of subscribers. S. Amudha et al. looked into the YouTube popular video metadata. The study employed an unsupervised dataset and the Decision Tree technique from Machine Learning to estimate the most effective courier service. The study used a views ratio per category to provide a simplified output of views, likes, dislikes, and comments scatter plot. Using pre-processing analysis, the thesis aids in understanding the value of these attributes (S. Amudha et al., 2020) [3].
Key Words: YouTube, Trending Section, Metadata Analysis, Python, Data Representation, Pandas, Plotly, Heroku Deployment
1. INTRODUCTION A particular video content on YouTube is not limited not to the actual video content being displayed to the user. A user visiting a YouTube video is encountered with likes, video title, description, comments, publishing date, publishing channel, number of subscribers of the channel and much more elements. All these features directly or indirectly contribute to a video being on the trending section.
3. DATA GATHERING We had to create a Python script to fetch the data from the YouTube Data API. The YouTube data API has a clean interface to obtain any type of data from their platform. It authenticates via API keys. The key was generated by registering an account under Google Cloud Platform (GCP), then creating a project and then enabling the YouTube API under that project. The current thresholds for this API are maximum of 10k calls per day which is more than enough for our project [4].
This dashboard can help in finding and comparing key aspects of a YouTube video and help budding creators, businesses to align their content to the most popularly used key aspects be included in the trending section in future.
The data returned by the YouTube API is of the nested JSON format that further needs data wrangling. Also, one issue we had to deal with was the paginated results. The JSON results were paginated with each page having token for the next one. Here is one of the JSON output for a video as show in Fig-1.
2. LITERATURE SURVEY YouTube is one of the most popular platforms and a lot of research has been done on it. Despite its importance, YouTube trending videos analysis has yet to be thoroughly investigated. Trending video analysis still has a lot of room for improvement. Ouyang, Li, and Li investigated the prediction of internet video popularity. The popularity forecasting problem was divided into two tasks in this study: video popularity prediction and video view count prediction. With
© 2022, IRJET
|
Impact Factor value: 7.529
|
ISO 9001:2008 Certified Journal
|
Page 1330