Date of Publication :7th March 2026

Abstract: The rapid development of social media platforms has changed the way content is being created, propagated and advertised. The launch of short form content such as Tik Toks has completely shifted the market and changed the way it functions. Earlier the same market was dominated by long form content and the way to get engagement and virality was completely different as to what it is now. 12 million shorts are uploaded to YouTube every single day and approximately 1 to 5 percent of them reach virality. This project, "Beyond Detection: An Integrated Machine Learning System for Predicting YouTube Virality," focuses on developing a smart system that helps predict what goes viral in the YouTube short space. The primary objective of the project is quantifying virality and engagement using features already known to the user. The aim is to build a system which can input the features commonly associated with YouTube shorts and use that information as well as the trend of historical data to predict how well a certain video will perform on the platform. The dataset used in this project has been preprocessed and cleaned to prepare for the necessary operations. I used a dataset collected by me using scraping to make it as accurate to my aim as possible. The data set consists of half the data acquired from YouTube shorts with the help of python libraries such as scrapy, Beautiful Soup etc. and half the data acquired from Instagram using a scraping software called Apify. Then the data is combined based on video title and preprocessed in Power BI.

Reference :